WO2022033377A1 - 一种媒体信息传输方法及电子设备 - Google Patents
一种媒体信息传输方法及电子设备 Download PDFInfo
- Publication number
- WO2022033377A1 WO2022033377A1 PCT/CN2021/110644 CN2021110644W WO2022033377A1 WO 2022033377 A1 WO2022033377 A1 WO 2022033377A1 CN 2021110644 W CN2021110644 W CN 2021110644W WO 2022033377 A1 WO2022033377 A1 WO 2022033377A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- message
- media information
- feature data
- transmission interface
- application
- Prior art date
Links
- 230000005540 biological transmission Effects 0.000 title claims abstract description 510
- 238000000034 method Methods 0.000 title claims abstract description 236
- 238000000605 extraction Methods 0.000 claims abstract description 181
- 238000012545 processing Methods 0.000 claims description 266
- 230000004044 response Effects 0.000 claims description 119
- 238000004891 communication Methods 0.000 claims description 103
- 230000008569 process Effects 0.000 claims description 62
- 238000004590 computer program Methods 0.000 claims description 22
- 230000000694 effects Effects 0.000 abstract description 10
- 238000013473 artificial intelligence Methods 0.000 description 394
- 238000012549 training Methods 0.000 description 89
- 230000006870 function Effects 0.000 description 79
- 238000013528 artificial neural network Methods 0.000 description 31
- 238000010586 diagram Methods 0.000 description 29
- 230000003993 interaction Effects 0.000 description 27
- 238000010295 mobile communication Methods 0.000 description 22
- 230000002457 bidirectional effect Effects 0.000 description 10
- 230000010365 information processing Effects 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 8
- 238000007781 pre-processing Methods 0.000 description 8
- 230000009467 reduction Effects 0.000 description 8
- 230000006835 compression Effects 0.000 description 7
- 238000007906 compression Methods 0.000 description 7
- 230000003287 optical effect Effects 0.000 description 7
- 238000005457 optimization Methods 0.000 description 7
- 230000001360 synchronised effect Effects 0.000 description 7
- 238000013475 authorization Methods 0.000 description 5
- 238000003062 neural network model Methods 0.000 description 5
- 238000012805 post-processing Methods 0.000 description 5
- 238000012790 confirmation Methods 0.000 description 4
- 235000019800 disodium phosphate Nutrition 0.000 description 4
- 239000013307 optical fiber Substances 0.000 description 4
- 230000005236 sound signal Effects 0.000 description 4
- 229920001621 AMOLED Polymers 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 3
- 238000004220 aggregation Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000013480 data collection Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 230000005855 radiation Effects 0.000 description 3
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013075 data extraction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 239000002096 quantum dot Substances 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 230000006837 decompression Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000003032 molecular docking Methods 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/38—Information transfer, e.g. on bus
- G06F13/42—Bus transfer protocol, e.g. handshake; Synchronisation
- G06F13/4265—Bus transfer protocol, e.g. handshake; Synchronisation on a point to point bus
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/38—Information transfer, e.g. on bus
- G06F13/382—Information transfer, e.g. on bus using universal interface adapter
- G06F13/385—Information transfer, e.g. on bus using universal interface adapter for adaptation of a particular data processing system to different peripheral devices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/38—Information transfer, e.g. on bus
- G06F13/40—Bus structure
- G06F13/4063—Device-to-bus coupling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/44—Program or device authentication
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/08—Network architectures or network communication protocols for network security for authentication of entities
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/20—Network architectures or network communication protocols for network security for managing network security; network security policies in general
- H04L63/205—Network architectures or network communication protocols for network security for managing network security; network security policies in general involving negotiation or determination of the one or more network security mechanisms to be used, e.g. by negotiation between the client and the server or between peers or by selection according to the capabilities of the entities involved
Definitions
- the embodiments of the present application relate to the field of media technologies, and in particular, to a method and electronic device for transmitting media information.
- image acquisition devices collect original video or image resources, compress the video or image resources into video streams, and transmit them to the corresponding Servers or devices that implement machine algorithm applications, these devices decode or decompress the received video streams to recover standard audio and video signals; then use machine algorithms such as deep learning to process audio and video signals to obtain artificial intelligence (artificial intelligence (artificial intelligence). , AI) application processing results.
- artificial intelligence artificial intelligence (artificial intelligence). , AI) application processing results.
- the present application provides a media information transmission method and electronic device, which are used to solve the problems in the prior art that a large bandwidth resource is occupied during the media data transmission process, which is not conducive to the subsequent application of the media resources.
- the present application provides a media information transmission method, which is applied to a first device, where the first device includes a first transmission interface, and the method includes: collecting first media information; Feature extraction, determining the first feature data of the first media information; sending the first feature data to the second device through the first transmission interface, and the first feature data is used by the second device to obtain the first feature data. the result of an application.
- the first application may be an AI application, and the device executing the above method may be the first device, or may be a module or component with corresponding functions, such as a chip.
- the following takes the first device performing the method as an example, and the first device may be a device having the capability of collecting media information and the capability of feature extraction.
- the first device may include a first transmission interface, and the first transmission interface supports the transmission of characteristic data, thereby realizing the transmission of media information to the second device by means of characteristic data, avoiding the need for media information at the transmitting end in the prior art. Compression coding is performed, and the media information transmitted needs to be restored at the receiving end of the media information.
- the coding and decoding process is relatively complex, which reduces the overhead of computing resources and reduces the overall delay of the system, which is helpful for AI applications with high real-time requirements. In applications, improve the user experience of using AI applications.
- the abstracted feature data is transmitted, the encoding difficulty is reduced, the amount of information transmitted is also greatly reduced, and the overhead of transmission resources is reduced. Further, considering the security when transmitting media information, when transmitting abstracted feature data, compared with the way of transmitting media information in the prior art, the feature data transmitted in the transmission interface cannot be reversely converted into original media information, so as to achieve more Good privacy protection capabilities.
- a possible implementation manner, before the sending the first feature data to the second device, further includes:
- the capability negotiation request message is used to request a transmission protocol supported by the first device and a feature extraction capability of the first device;
- the transmission protocol of the first device is used to instruct the first device to support the transmission of feature data;
- the feature extraction capability of the first device is used to instruct the first device to support extraction of the first feature of the first media information data; send a capability negotiation response message to the second device through the first transmission interface;
- the capability negotiation response message is used to confirm that the first device supports the transmission protocol for transmitting feature data and the features of the first device extraction capacity.
- a possible implementation manner in response to the first operation on the first application, sending a first notification message to the second device;
- the first device is an electronic device that establishes a communication connection with the second device;
- the first notification message is used to request the first device to establish a first application collaboration with the second device;
- the first response message is used to confirm the The first device and the second device start the first application collaboration.
- the first operation on the first application of the first device can trigger the process of establishing the first application collaboration between the first device and the second device, and confirm whether to enable the first application collaboration through the second device.
- the first application collaboration may be a collaborative process in which the first device sends the first characteristic data to the second device through the first transmission interface, and the second device obtains the result of the first application according to the first characteristic data, so as to improve the user's use of the first application.
- a device and a second device cooperate to process the experience of the first application.
- a possible implementation manner receiving a first notification message sent from the second device; the first device is an electronic device that establishes a communication connection with the second device; the first notification message is used to request the The first device and the second device establish a first application collaboration; in response to a third operation on the first application, send a first response message to the second device; the first response message is used to confirm the The first device and the second device start the first application collaboration.
- the second device triggers the first device and the second device to establish the first application collaboration establishment process, and according to the third operation of the first application on the first device, confirms whether to enable the first application collaboration, and enhances the user experience.
- the experience of the first application is co-processed using the first device and the second device.
- a possible implementation manner, before the sending the first feature data to the second device, further includes:
- a capability of the result of an application; a capability negotiation response message from the second device is received through the first transmission interface; the capability negotiation response message is used to confirm that the second device supports the transmission protocol for transmitting feature data and all Describe the feature data processing capability of the second device.
- the capability negotiation method for example, the first device initiates a negotiation request message or the second device initiates a negotiation request message, thereby confirming whether the first device has the feature extraction capability and the capability of transmitting feature data, and the first device Whether the second device supports the ability to receive feature data and the ability to process feature data, so that the first device and the second device can confirm whether they support the transmission of feature data and cooperate to implement the corresponding functions of the first application to improve the performance of the AI application.
- the method before the first device performs feature extraction on the first media information, the method further includes:
- the first feature extraction model is used to perform feature extraction on the first media information, and the version of the first feature extraction model corresponds to the version of the first feature data processing model, so
- the first feature data processing model is used by the second device to process the first feature data to obtain a result of the first application.
- the first device and the second device After the first device and the second device establish a connection and determine the task of the first application that can execute the corresponding collaborative processing, the first device and the second device respectively load the algorithm model corresponding to the first application. Input part (first feature data extraction model) and output part (first feature data processing model). Thus, the first application collaboration of the first device and the second device is realized.
- the capability negotiation response message further includes:
- the version of the feature extraction model in the first device or, the version of the feature data processing model in the second device.
- the obtaining the first feature extraction model includes:
- the first feature extraction model from the second device is received through the first transmission interface, or the first feature extraction model is received from a server, or the first feature extraction model stored in the first device is read.
- the first feature extraction model can be obtained in various ways, so as to realize more flexible ways of AI application collaboration.
- the method further includes:
- the version of the first feature extraction model corresponds to the version of the first feature data processing model, and the first feature data processing model is used by the second device to process the first feature data to obtain the The results of the first application are described.
- the first device may store the first feature data processing model, and through the above method, the first feature data processing model is sent to the second device through the first transmission interface, so that the first device and the second device can implement AI Applications co-process media information.
- the method further includes:
- the version of the second feature extraction model corresponds to the version of the second feature data processing model
- the second feature extraction model and the second feature data processing model are to update the first feature Determined after the extraction model and the second feature data processing model.
- the updated feature extraction model ie, the second feature extraction model
- the needs of various AI applications can be adapted, and the first device and the second device can be improved.
- the method further includes:
- feature extraction is performed on the training samples to generate first training feature data
- the first training feature data is sent to the second device through the first transmission interface; the first training feature data is used to train the first feature extraction model and the first feature data processing model.
- the first device and the second device can be used for joint training, and the computing power of the first device and the second device can be reasonably utilized to improve the performance of the AI application.
- feedback data from the second device is received through the first transmission interface, where the feedback data is determined by the second device after training according to the first training feature data; the The feedback data is used by the first device to train the first feature extraction model.
- the feature extraction model on the first device is trained through the feedback data fed back by the second device, and the performance of the feature extraction model is improved by using the training effect of joint training of the first device and the second device.
- the method further includes:
- the first device can adjust the state of media information collection by the first device according to the first message sent by the second device, so as to better obtain the media information collected by the first application and improve the effect of the first application.
- the state in which the first device collects media information includes at least one of the following: an on state, an off state, or a parameter for collecting media information.
- the method further includes: receiving a second message from the second device through the first transmission interface; wherein the second message is used to instruct the first device to obtain the first data; obtaining the first data in response to the second message, or collecting the first data; sending the first data to the second device; the first data is one of the following: the Media information collected by the first device, parameters of the first device, data stored by the first device, and data received by the first device.
- the second device can instruct the first device to obtain the first data, so that the transmission of the characteristic data between the first device and the second device is compatible with the transmission of other data, which helps to improve the first device and the second device.
- the first data is sent to the second device through the first transmission interface.
- the first transmission interface can support the transmission of various data, so as to improve the transmission performance and the adaptability of transmission scenarios.
- a possible implementation manner receiving a second message from the second device through the first transmission interface, where the second message is used to instruct the first device to collect feature data of the third media information; in response to For the second message, collect the third media information; perform feature extraction on the third media information to obtain third feature data; and send the third feature to the second device through the first transmission interface data.
- the second message can be sent by the second device to control the first device to collect media information and transmit the corresponding third feature data.
- it can be determined by the processing result of the first application or the needs of the AI application
- the third media information to be collected can flexibly adjust the media information collected by the first device, so that the AI application can obtain better results, thereby improving the effect of the AI application as a whole.
- the second message or the first message is determined by the second device according to the first feature data.
- the second device can determine the result of the first application based on the first feature data transmitted by the first device, and generate the first message or the second message according to the result of the first application, so as to feed back the corresponding information to the first device.
- the first device can adjust the collection, acquisition and transmission of media information in response to the first message or the second message, so that the first device and the second device can better complete the first application collaboration.
- the first device may further include a display unit; the method further includes: receiving a third message from the second device through the first transmission interface, where the third message is the Determined by the second device according to the first feature data, the third message is used to indicate the content displayed by the first device; in response to the third message, the display unit displays the content used in the third message. to instruct the first device to display content.
- the content to be displayed can be obtained through the third message, and the content can be the processing result of the first application, or the content that other second devices need to display by the first device.
- the two devices can better realize the collaboration of AI applications and improve the experience of using AI applications.
- the method further includes:
- An authentication request message sent by the second device is received through the first transmission interface, where the authentication request message is used to request whether the first device establishes a communication connection with the second device, and the communication connection is used to confirm the permission of the second device to control the first device;
- the authentication response message is used to confirm the authority of the second device to control the first device.
- the second device can obtain the authority to control the first device through the authentication between the first device and the second device, so that after the second device obtains the result of the first application according to the first feature data , and adjusting the media information collected by the first device is helpful for obtaining better results of the first application and improving the performance of the AI application.
- the method further includes:
- An authentication success message sent by the second device is received through the first transmission interface; the authentication success message includes: the device identifier corresponding to the first device, and the location where the first device and the second device are located. The identity of the distributed system.
- the first device and the second device can also be set as devices in a distributed system, so as to achieve better management of the first device and the second device, and it is beneficial to use multiple devices to achieve AI application collaboration.
- the first device includes a first module; the authentication success message further includes at least one of the following: an identifier of the first module of the first device, and the first module in the Identity in a distributed system.
- the modules in the first device can also be set as modules in the distributed system, thereby preparing for the second device to control the modules in each device and to cooperate to complete the AI application.
- the first device and the second device establish a channel connection through a third transmission interface; the feature data or message sent by the first device is encapsulated as the first device through the first transmission interface. After bit stream data, it is sent through the third transmission interface.
- the feature data is encapsulated through the first transmission interface, and the encapsulated data is sent to the second device through the third transmission interface, so that through the third transmission interface, multiple transmission protocols can be compatible, and aggregated transmission can also be realized. function, thereby improving the transmission capability of the first device and the compatibility of transmitting media information.
- the first device and the second device establish a channel connection through a third transmission interface; the message received by the first device is the second bit stream data received through the third transmission interface , and obtained by decapsulating the second bit stream data through the first transmission interface.
- the data from the second device received by the third transmission interface is decapsulated through the first transmission interface, and the third transmission interface can be compatible with multiple transmission protocols, and can also realize functions such as aggregated transmission, thereby improving the first device. transmission capability and compatibility of transmission media information.
- the present application provides a method for transmitting media information, which is applied to a second device; the second device includes a second transmission interface; the method includes: receiving a first transmission interface from the first device through the second transmission interface One feature data; the first feature data is determined according to the feature extraction of the first media information collected by the first device; the first feature data is processed to obtain a processing result of the first application.
- the first application may be an AI application
- the device that executes the above method may be a second device, or a module or component with corresponding functions, such as a chip.
- the following takes the second device performing the method as an example, and the second device may be a device having the capability of receiving feature data and the capability of processing feature data.
- the second device may include a second transmission interface, where the second transmission interface supports the transmission of feature data, thereby enabling the second device to receive feature data instead of directly receiving media data, avoiding the need for media information at the sending end in the prior art Compression coding is performed, and the media information transmitted needs to be restored at the receiving end of the media information.
- the coding and decoding process is relatively complex, which reduces the overhead of computing resources and reduces the overall delay of the system, which is helpful for AI applications with high real-time requirements. In the application, improve the user's experience of using AI applications.
- the abstracted feature data is transmitted, the encoding difficulty is reduced, the amount of information transmitted is also greatly reduced, and the overhead of transmission resources is reduced. Further, considering the security when transmitting media information, when transmitting abstracted feature data, compared with the way of transmitting media information in the prior art, the feature data transmitted in the transmission interface cannot be reversely converted into original media information, so as to achieve more Good privacy protection capabilities.
- a possible implementation manner in response to a second operation on the first application, sending a first notification message to the first device;
- the first device is an electronic device that establishes a communication connection with the second device;
- the first notification message is used to request the first device to establish a first application collaboration with the second device;
- the first response message is used to confirm the The first device and the second device start the first application collaboration.
- the second device responds to the second operation on the first application, triggers the first device and the second device to establish the first application collaboration establishment process, and confirms through the first device whether to enable the first application collaboration,
- the user's experience of co-processing the first application using the first device and the second device is improved.
- a possible implementation is to receive a first notification message sent by a first device; the first device is an electronic device that establishes a communication connection with the second device; the first notification message is used to request the first notification message The device establishes a first application collaboration with the second device; in response to a fourth operation on the first application, a first response message is sent to the first device; the first response message is used to confirm the first The device starts the first application collaboration with the second device.
- the first device can trigger the first device and the second device to establish the first application collaboration establishment process, and the second device can respond to the fourth operation on the first application to confirm whether to enable the first application collaboration.
- the first application collaboration may be a collaborative process in which the first device sends the first characteristic data to the second device through the first transmission interface, and the second device obtains the result of the first application according to the first characteristic data, so as to improve the user's use of the first application.
- a device and a second device cooperate to process the experience of the first application.
- the capability negotiation request message is used to request the transmission protocol supported by the first device and the feature extraction capability of the first device;
- the transmission protocol of the first device is used to instruct the first device to support the transmission of feature data;
- the feature extraction capability of the first device is used to instruct the first device to support extraction of the first feature data of the first media information ;
- the capability negotiation response message is used to confirm that the first device supports a transmission protocol for transmitting characteristic data.
- a capability negotiation request message sent by the first device is received through the second transmission interface; the capability negotiation request message is used to request a transmission protocol supported by the second transmission interface and feature data processing of the second device capability, the transmission protocol of the second device is used to instruct the second device to support the transmission of feature data; the feature data processing capability of the second device is used to instruct the second device to support processing the acquisition of the first feature data the capability of the result of the first application; send a capability negotiation response message to the first device through the second transmission interface; the capability negotiation response message is used to confirm that the second device supports the transmission protocol and all Describe the feature data processing capability of the second device.
- a possible implementation manner, before the receiving the first feature data, further includes:
- the first feature data processing model is used by the second device to process the first feature data to obtain the result of the first application; the version of the first feature extraction model is the same as the corresponds to the version of the first feature data processing model, and the first feature extraction model is used to perform feature extraction on the first media information.
- the first device and the second device After the first device and the second device establish a connection and determine the task of the first application that can execute the corresponding collaborative processing, the first device and the second device respectively load the algorithm model corresponding to the first application. Input part (first feature data extraction model) and output part (first feature data processing model). Thus, the first application collaboration of the first device and the second device is realized.
- the capability negotiation response message further includes: the version of the feature extraction model in the first device; or the version of the feature data processing model in the second device.
- the acquiring the first feature data processing model includes: receiving the first feature data processing model from the first device through the second transmission interface, or receiving the first feature data from a server a data processing model, or, reading the first feature data processing model stored in the second device;
- the method further includes: sending a first feature extraction model to a first device through the second transmission interface; wherein the version of the first feature extraction model is processed with the first feature data The version of the model corresponds, and the first feature data processing model is used by the second device to process the first feature data to obtain the result of the first application.
- the method further includes: acquiring a second feature data processing model, the version of the second feature data processing model corresponds to the version of the second feature extraction model, the second feature extraction model and the second feature data processing model is determined after updating the first feature extraction model and the second feature data processing model.
- the first device can obtain the first feature extraction model and the second device can obtain the first feature data processing model in various ways, so as to realize more flexible ways of AI application collaboration.
- the method further includes: receiving first training feature data; the first training feature data is determined after the first device performs feature extraction on a training sample according to the first feature extraction model ; Train the first feature data processing model according to the first training feature data.
- the method further includes: obtaining feedback data of a first feature extraction model; the feedback data is determined by the second device after training according to the first training feature data; the feedback data for the first device to train the first feature extraction model; and to send the feedback data to the first device.
- the method further includes: receiving second feature data sent by the second device; the second feature data is the second media information collected by the first device and the second The feature extraction model is determined after feature extraction; the second feature data is processed according to the second feature data processing model to obtain the result of the first application.
- the first device and the second device can be used for joint training, and the computing power of the first device and the second device can be reasonably utilized to improve the performance of the AI application.
- the method further includes: sending a first message to the first device through the second transmission interface; the first message is used to indicate the state of media information collection by the first device.
- the second device can send the first message to the first device to adjust the state of media information collection by the first device, so as to better obtain the media information collected by the first application and improve the effect of the first application.
- the state in which the first device collects media information includes at least one of the following: an on state, an off state, or a parameter for collecting media information.
- the method further includes: sending a second message to the first device through the second transmission interface; the second message is used to instruct the first device to acquire the first data;
- the first data is one of the following: media information collected by the first device, parameters of the first device, data stored by the first device, and data received by the first device.
- the second device can instruct the first device to obtain the first data, so that the transmission of the characteristic data between the first device and the second device is compatible with the transmission of other data, which helps to improve the first device and the second device.
- the method further includes: receiving the first data from the first device through the second transmission interface.
- the second transmission interface can support the transmission of various data, so as to improve the transmission performance and the adaptability of transmission scenarios.
- the method further includes: sending a second message to the first device through the second transmission interface; the second message is used to instruct the first device to collect feature data of the third media information ; Receive third feature data sent from the first device through the second transmission interface; the third feature data is determined after the first device performs feature extraction on the collected third media information.
- the second message can be sent by the second device to control the first device to collect media information and transmit the corresponding third feature data.
- it can be determined by the processing result of the first application or the needs of the AI application
- the third media information to be collected can flexibly adjust the media information collected by the first device, so that the AI application can obtain better results, thereby improving the effect of the AI application as a whole.
- the first message or the second message is determined according to a processing result of the first feature data.
- the second device can determine the result of the first application based on the first feature data transmitted by the first device, and generate the first message or the second message according to the result of the first application, so as to feed back the corresponding information to the first device.
- the first device can adjust the collection, acquisition and transmission of media information in response to the first message or the second message, so that the first device and the second device can better complete the first application collaboration.
- the first device further includes a display unit; the method further includes:
- a third message is generated; the third message is used to indicate the content displayed by the first device.
- the content to be displayed can be obtained through the third message, and the content can be the processing result of the first application, or the content that other second devices need to display by the first device.
- the two devices can better realize the collaboration of AI applications and improve the experience of using AI applications.
- the number of the first devices is N; the method further includes:
- a fourth message is received through the second transmission interface; the fourth message includes M first characteristic data of the N first devices; N and M are positive integers greater than 1; M is greater than or equal to N;
- the M pieces of first feature data are processed to obtain the result of the first application.
- the method further includes:
- the second device controls the authority of the first device; the authentication response message sent by the second device is received through the second transmission interface; the authentication response message is used to confirm whether the first device is compatible with the first device.
- the two devices establish a communication connection.
- the second device can obtain the authority to control the first device through the authentication between the first device and the second device, so that after the second device obtains the result of the first application according to the first feature data , adjusting the media information collected by the first device is helpful to obtain better results of the first application and improve the performance of the AI application.
- the method further includes: in response to an authentication response message sent by the second device, setting a device identity corresponding to the first device for the first device, and the first device and the The identifier of the distributed system where the second device is located; the device identifier corresponding to the first device and the identifier of the distributed system are used for the first device to communicate with the second device; the first device The device sends an authentication success message to the second device through the first transmission interface; the authentication success message includes: the device identifier corresponding to the first device, and the location where the first device and the second device are located. The identity of the distributed system.
- the first device and the second device can also be set as devices in a distributed system, so as to achieve better management of the first device and the second device, and it is beneficial to use multiple devices to achieve AI application collaboration.
- the second device includes a second module; the authentication success message further includes at least one of the following: an identifier of the second module, and the second module in the distributed system 's identification.
- the modules in the first device can also be set as modules in the distributed system, thereby preparing for the second device to control the modules in each first device and to cooperate to complete the AI application.
- the second device further includes a third transmission interface; the first device and the second device establish a channel connection through the third transmission interface; the message sent by the second device is through the third transmission interface. After the second transmission interface is encapsulated into the second bit stream data, the data is sent through the third transmission interface.
- the data is encapsulated through the second transmission interface, and the encapsulated data is sent to the first device through the third transmission interface, so that through the third transmission interface, multiple transmission protocols can be compatible, and functions such as aggregation transmission can also be realized. , thereby improving the transmission capability of the second device and the compatibility of transmitting media information.
- the first device and the second device establish a channel connection through a third transmission interface; the feature data or message received by the second device is the first device received through the third transmission interface. bit stream data, and obtained by decapsulating the second bit stream data through the second transmission interface.
- the data from the first device received by the third transmission interface is decapsulated through the second transmission interface, and the third transmission interface can be compatible with multiple transmission protocols, and can also realize functions such as aggregated transmission, thereby improving the second device. transmission capability and compatibility of transmission media information.
- the present application provides an electronic device, the electronic device includes a memory and one or more processors; wherein the memory is used to store computer program code, and the computer program code includes computer instructions; when the The computer instructions, when executed by the processor, cause the electronic device to perform the method of any one of the first aspects.
- the present application provides an electronic device, the electronic device includes a memory and one or more processors; wherein, the memory is used to store computer program code, and the computer program code includes computer instructions; when the The computer instructions, when executed by the processor, cause the electronic device to perform the method in any of the possible implementations of the first aspect or the second aspect.
- the present application provides a media information transmission system, including: the electronic device described in the third aspect and the electronic device described in the fourth aspect.
- the present application provides a computer-readable storage medium, the computer-readable storage medium comprising program instructions, when the program instructions are executed on an electronic device, the electronic device is made to perform any one of the first aspects. Any possible method of the second aspect, or the electronic device is caused to perform any one of the possible methods of the second aspect.
- FIG. 1a is a schematic structural diagram of a media information sending device in the prior art
- FIG. 1b is a schematic structural diagram of a media information receiving device in the prior art
- 1c is a schematic structural diagram of a media information receiving device in the prior art
- 2a is a schematic diagram of a media information transmission method provided by the application
- 2b is a schematic structural diagram of an AI algorithm model provided by the application.
- 3a is a schematic structural diagram of a first device provided by the application.
- 3b is a schematic structural diagram of a second device provided by the application.
- 3c is a schematic diagram of a distributed system architecture provided by the application.
- 4a is a schematic flowchart of a method for establishing a communication connection for AI application collaboration provided by the application;
- 4b-4c are schematic diagrams of a search interface of a first device provided by the present application.
- 4d-4e are schematic interface diagrams of an AI application collaboration provided by the present application.
- 5a is a schematic diagram of a distributed system architecture provided by the application.
- 5b is a schematic flowchart of a media information transmission method provided by the application.
- FIG. 5c is a schematic diagram of a scenario provided by this application.
- FIG. 5d is a schematic diagram of an AI application provided by this application.
- 6a is a schematic diagram of a distributed system architecture provided by the application.
- 6b is a schematic flowchart of a media information transmission method provided by the application.
- FIG. 6c is a schematic diagram of a scenario provided by this application.
- FIG. 6d is a schematic diagram of an AI application provided by this application.
- 7a is a schematic diagram of a distributed system architecture provided by the application.
- 7b is a schematic flowchart of a media information transmission method provided by the application.
- FIG. 7c is a schematic diagram of a scenario provided by this application.
- FIG. 7d is a schematic diagram of an AI application provided by this application.
- FIG. 8 is a schematic structural diagram of a possible electronic device according to an embodiment of the present application.
- FIG. 9 is a schematic structural diagram of another possible electronic device according to an embodiment of the present application.
- the media information involved in this application may include: image information, audio information, video information, sensor information, and other media information collected by the first device.
- the collected video image information may be audio, video, visible light images, or information such as radar and depth.
- the first device may include: a camera, a sensor, a microphone, or other devices or units with a function of collecting media information.
- the media information acquired by the first device involved in this embodiment of the present application may be an original image, for example, may be an output image of a camera, that is, the camera converts the light information reflected by the collected object into a digital image.
- the raw data obtained from the signal, the raw data has not been processed.
- the original image may be raw format data.
- the raw format data may include object information and camera parameters.
- the camera parameters may include sensitivity (international standardization organization, ISO), shutter speed, aperture value, white balance, and the like.
- the original image may also be the input image of the ISP, or the original image may also be the input image of a network neural unit, such as the neural-network processing unit (NPU) below.
- the output image of the network neural unit may be a high dynamic range image (high dynamic range, HDR) image, or may be another processed image, which is not limited herein.
- the media information acquired by the first device involved in the embodiment of the present application may also be an output image of an ISP.
- the ISP processes the original image to obtain an image in RGB format or YUV format, and adjusts the brightness of the image in RGB format or YUV format. image obtained after.
- the specific value to which the ISP adjusts the brightness of the image in the RGB format or the YUV format may be set by the user, or may be set by the mobile phone when it leaves the factory.
- the media information acquired by the first device may also be an input image of a processor such as a graphics processing unit (graphics processing unit, GPU) of the first device hereinafter.
- the "media information” involved in the embodiments of the present application such as original media information, media information obtained by the first device, media information processed by the second device (for example, HDR images), etc., when the media information is an image
- the media information can refer to a picture, or it can be a set of some parameters (for example, pixel information, color information, brightness information).
- the multiple involved in the embodiments of the present application refers to greater than or equal to two.
- the sending end device of the media information involved in the embodiments of the present application may be a device having a function of collecting media information.
- the media information may include one or more of image information, video information, audio information, and sensor information.
- the sending end device of the media information may be a unit or device with a video image collection function, and the collected video image information may be audio, video, visible light images, or media information such as radar and depth.
- the sending end device of the media information may include a video collection unit such as a camera for collecting video information or image information, and may also include an audio collection unit such as a microphone for collecting audio information.
- the video capture unit may be one or more of units such as an optical lens, an image sensor, a microphone, etc., and is used to capture the original media frequency signal (audio, image or mix).
- the sending end device of media information may be: mobile terminal such as mobile phone tablet, smart home terminal such as smart TV, AR/VR head-mounted display, vehicle camera, external camera and other mobile phone accessories.
- the sending end device of the media information may be a terminal device including a media collection unit, such as a smart screen.
- the sending end device of the media information collects the original audio and video information, and after processing, forms an audio and video signal in a standard format.
- the sending end device of the media information can also be used as the sending end of the media information, which is sent to the receiving end through the transmission interface or the network after being compressed and encoded.
- the transmission interface may be a media transmission interface such as HDMI, DP, and USB.
- the sending end device of the media information may also be a device that obtains media information and sends the media information.
- the sending end device of the media information may obtain the media information from the network or from a local storage unit, and Send the media information to the second device.
- the sending end device of the media information may not be a device having the function of collecting the media information, that is, the sending end device of the media information may only be a device that sends the sending function of the media information.
- the sending end device of the media information can send the obtained audio and video media information to the receiving end device through the transmission interface or the network after being compressed and encoded.
- the transmission interface may be a media transmission interface such as HDMI, DP, and USB.
- the media information sending end device shown in FIG. 1a may include: a media signal collection unit (for example, an audio collection unit) unit and video capture unit), media encoding unit (eg, audio encoder and video encoder) and output interfaces (eg, audio output interface and video output interface).
- the media signal acquisition unit may have multiple implementation forms.
- the media signal acquisition unit may include at least one of the following: a media signal acquisition unit, a media signal receiving unit, and a storage unit.
- the media signal collection unit can be used to collect original media signals.
- it may include one or more of media signal acquisition sensors or devices such as optical lenses, image sensors, microphones, and radars.
- the acquired media information may be audio, video, visible light images, or information such as radar and depth.
- the media signal receiving unit can be used to receive media signals from the network or other devices.
- the storage unit can be used to store media signals locally in the sending end device, and of course, can also be used to store other information.
- the media coding unit is used to perform media coding and channel coding on the media signal acquired by the first device according to the media coding protocol, the link layer and the physical layer protocol to obtain the physical layer signal, and transmit the physical layer signal to the output interface, thereby sending it to the corresponding The receiver device of the media information.
- the apparatus for acquiring media information may further include a media signal preprocessing unit.
- the media information preprocessing unit can be used to perform preprocessing such as noise reduction and restoration on the original audio and video media signals.
- the video preprocessing unit can be used to perform preprocessing such as noise reduction and demosaicing on the original video frame image.
- the receiving end device of the media information may be a media processing device.
- the receiving end device may be a terminal device such as a mobile phone, a tablet computer, a smart TV, and a vehicle-mounted computer.
- Electronic devices may also be referred to as terminal devices.
- Terminal equipment may also be referred to as user equipment, access terminal, subscriber unit, subscriber station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent, or user equipment.
- the terminal device can be a mobile phone (mobile phone), a tablet computer (pad), a computer with a wireless transceiver function, a virtual reality (VR) terminal, an augmented reality (AR) terminal, an industrial control (industrial control) terminal wireless terminal in self-driving, wireless terminal in remote medical, wireless terminal in smart grid, wireless terminal in transportation safety, wireless terminal in smart grid Wireless terminals in a smart city, wireless terminals in a smart home, etc.
- the receiving end device of the media information may also be a set-top box, a separate docking station (DOCK), a smart TV, a smart large screen, a mobile phone, a tablet computer or a personal computer (PC), a smart screen, a mobile phone, a smart camera, a smart Terminal equipment such as speakers and headphones.
- the smart screen can be an audio-visual entertainment center in the family, as well as an information sharing center, a control management center, and a multi-device interaction center.
- the terminal device may also be a portable terminal including functions such as a personal digital assistant and/or a music player, such as a mobile phone, a tablet computer, a wearable device (such as a smart watch) with a wireless communication function, a vehicle-mounted device, and the like.
- portable terminals include but are not limited to carrying Or portable terminals of other operating systems.
- the aforementioned portable terminal may also be, for example, a laptop computer (Laptop) having a touch-sensitive surface (eg, a touch panel).
- the above-mentioned terminal may also be a desktop computer having a touch-sensitive surface (eg, a touch panel).
- the receiver device of the media information may also be a processor chip in a set-top box, a display screen, a smart large screen, a television (television, TV), a mobile phone, or other terminal devices with a media information processing function.
- the chip device may be a system on chip (system on chip, SoC) or a baseband chip.
- SoC system on chip
- the second device may also be a computing device deployed with a graphics processing unit (graphics processing unit, GPU), a distributed computing device, or the like.
- the media processing apparatus may include: an input interface, a media decoding unit and a media information processing unit.
- the input interface may be used to receive a media signal sent from a sender (eg, a sender device of media information).
- the input interface is used for receiving the physical layer signal from the transmission channel, and the media decoding unit is used for decoding the media data signal from the physical layer signal according to the link layer and the physical layer protocol.
- the media decoding unit and the media information processing unit may include: a parser, an audio decoder, a video decoder, and a video post-processing unit.
- each unit can be realized by hardware, also can be realized by software, or can be realized by hardware combined with software.
- video decoders, video post-processing units, etc. are implemented by hardware logic
- AI analysis of media data, display strategy processing and other units can be implemented by software codes running on hardware processors
- other units such as audio decoders can be implemented by software accomplish.
- a possible scenario is that the receiving end device of the media information decompresses and decodes the encoded signal obtained from the interface channel and restores the audio and video media information.
- media files in formats such as mp4 are parsed by the parser.
- the audio encoding file may be audio elementary stream (elementary stream, ES) data
- the video encoding file may be video ES data.
- Audio encoded files are decoded by an audio decoder to obtain audio data; video encoded files are processed by a video decoder to obtain video frames.
- the receiving end device of the media information can also be used to synchronize the image obtained by the video post-processing with the audio data, so that the output of the audio output interface is synchronized with the output of the video output interface, that is, the audio output from the audio output interface is synchronized with the video output interface.
- the output video images are synchronized.
- the receiving end device of the media information may further include a display unit, and at this time, the received audio and video media information may be processed, and the audio and video media information may be played.
- the display unit may also be located in another device, and the device is a device that establishes a wireless or wired communication connection with the media data processing apparatus.
- the display unit may be located in a terminal device such as a display (or called a display screen), a television, and a projector.
- the display unit can be used to play media files processed by the data processing device, and can also play other media files.
- AI artificial intelligence
- the data processing device can perform AI applications and other processing on multimedia files, namely
- the data processing device may also have an AI processing capability for implementing the functions of corresponding AI applications.
- the receiving end decompresses and decodes the information set obtained from the interface channel and restores the audio and video media information, it can also perform image processing and other operations, and use the receiving end's AI software and hardware processing system to perform subsequent AI application processing , to obtain AI analysis results of media data.
- the receiving end device of the media information may further include: a neural network training unit and a neural network processing unit.
- the neural network training unit is used to train AI algorithm models with labeled data.
- the training process of the AI algorithm model can be performed offline on the server, or online on the device or cloud.
- the neural network processing unit is used to load one or more AI algorithm models, process the media data, and obtain the inference results of the AI algorithm models.
- the deep learning algorithm of the receiving end device of the media information and AI software and hardware systems such as NPU are used to perform AI processing on audio and video signals to obtain inference results such as detection, classification, identification, positioning, and tracking.
- the inference results can be used for The corresponding AI application scenarios. For example, use the inference results to realize functions such as biometric recognition, environment recognition, scene modeling, machine vision interaction, and voice interaction.
- the transmission involved in the embodiments of this application includes receiving and/or sending.
- the sending end of the media information and the receiving end of the media information can be connected by wired or wireless means and transmit the media information.
- the form of the transmission interface may be an electrical signal transmitted by wire, an optical signal transmitted by an optical fiber, a radio signal, a wireless optical signal, and the like.
- a physical channel connection may be established between the sending end device of the media information and the receiving end device of the media information through wired and/or wireless communication protocols such as copper wire, optical fiber, etc.
- FIG. 1c it is a schematic structural diagram of a network system for media information transmission according to an embodiment of the present application.
- the network system includes a media information sending end device and a media information receiving end device.
- a physical layer signal for transmitting media information may be transmitted through a transmission channel.
- the transmission channel may be a physical transmission channel such as copper wire and optical fiber.
- the signal of the transmitted media information may be a wired electrical signal, an optical signal, or the like.
- the data signal for transmitting the media information may be the data signal of the HDMI protocol, the data signal of the DP protocol, or the data signal of other protocols.
- the interface standards used by electronic devices to transmit media data include: high definition multimedia interface (high definition multimedia interface, HDMI), USB interface, DP interface, and the like.
- HDMI is an interface that transmits uncompressed digital high-definition multimedia (video and/or audio).
- HDMI uses transition minimized differential signaling (TMDS) technology.
- USB is a serial bus standard and a technical specification for input and output interfaces.
- the USB Type-C interface can support PD and support the transmission of other data than multimedia data.
- a possible way is to encode the audio and video media information obtained on the sending end of the media information before transmitting the media information, and then transmit the audio and video information to the media information after the audio and video encoding. the receiving end.
- This type of method transmits media signals in the channel.
- the media signals have a large amount of data during the transmission process, consume a lot of computing resources, have high costs, and have a large overall system delay, which is not conducive to AI applications that require high real-time performance.
- the media information needs to be compressed and encoded at the sending end, and the transmitted media information needs to be restored at the receiving end of the media information, the encoding and decoding process is more complicated, and it needs to consume more computing resources, resulting in a large overall system delay. It is beneficial to be used in AI applications with high real-time requirements.
- the present application provides a schematic flowchart of a method for transmitting media information.
- a corresponding algorithm model is correspondingly set.
- the pre-trained algorithm model can be Or the algorithm model obtained by online training is divided into two parts: the input part and the output part.
- the input part may be the first feature extraction model.
- the output part may be a feature data processing model for post-processing the feature data.
- the first feature extraction model may be a feature extraction unit of a convolutional neural network, and the first feature data processing model may be any neural network model (such as a convolutional neural network model), and may also be other algorithm models, This is not limited.
- Figure 2b takes the convolutional neural network model as an example for illustration.
- the input part is a feature extraction part (also called a feature extraction model), and the input part can include the input layer of the AI algorithm model, which is used to perform feature extraction on the acquired media information such as audio and video to obtain feature data.
- the output feature data may be feature data processed by convolution and weighting of the input layer of the AI algorithm model.
- x1 ⁇ x6 in Figure 2b are the convolution modules in the input layer.
- the output part is the part that performs post-processing on the feature data (it can also be called the feature data processing model).
- the output part includes the hidden layer and output layer of the model.
- the feature data after feature extraction by the product module is input to the convolution module in the hidden layer, and the obtained data is input to the output layer.
- the output layer can be a classifier to obtain the processing results of the AI algorithm model.
- the input part of the AI algorithm model corresponding to the AI application can be loaded into the first device, and the output part of the AI algorithm model corresponding to the AI application can be loaded into the second device.
- Both the first device and the second device have software and hardware capabilities required for AI applications, for example, including NPU and other computing hardware for processing the input part of the AI algorithm model and the AI algorithm model for AI application collaboration.
- the processor for processing the AI algorithm model as an NPU as an example, in this case, the input part can be deployed in the NPU of the first device, and the output part can be deployed in the NPU of the second device.
- Step 201 The first device acquires first media information.
- the first device acquires the first media information
- Step 202 The first device performs feature extraction on the first media information to determine first feature data of the first media data.
- the first device may determine the input part (the first feature extraction model) of the AI algorithm model corresponding to the AI application based on the AI application in the AI application collaboration between the first device and the second device, and load the first device on the first device.
- the input part of the AI algorithm model of the AI application performs feature extraction on the first media information.
- Step 203 The first device sends the first feature data to the second device.
- Step 204 The second device processes the first feature data to determine the result of the first application.
- the feature data output by the input part of the AI algorithm model loaded by the first device is transmitted to the second device through the transmission interface in this application, and the output part (the first feature data processing model) of the AI algorithm model loaded by the second device is used. ) to obtain the final output of the AI model after further processing.
- the media information acquired by the first device may not be transmitted to the second device, but the NPU of the first device is used to run the input part of the AI algorithm model to convert the media information into feature data and then transmit it.
- the data transmission interface transmits the feature data to the second device for processing in the output part of the AI model.
- the processing capability of the feature extraction model is added to the first device that sends the media information, so that the sender does not need to send the media information, but sends the feature data after feature extraction of the media information. Since the data volume of feature data is significantly lower than the original audio and video information, and additional compression and encoding process for media information can be omitted, system power consumption and delay can be reduced, and costs can be reduced. Real-time transmission under bandwidth conditions improves product competitiveness. In addition, since the characteristic data transmitted in the transmission interface cannot be reversely converted into the original media information, better privacy protection capability is achieved.
- the input part and the output part of the AI algorithm model can be stored in the first device and the second device respectively, or both can be stored in the first device, the second device or the cloud server. , which is not limited here.
- the first device and the second device After the first device and the second device establish a connection and determine that the corresponding AI task of co-processing can be performed, the first device and the second device respectively load the input part and the output part of the algorithm model corresponding to the AI application, and confirm the first device and the second device respectively. Whether the first device and the second device successfully load the input part and the output part of the AI application algorithm model.
- the second device confirms that the first device fails to load the input part of the algorithm model corresponding to the AI application
- the second device confirms whether the first device has the ability to load the input part of the algorithm model for processing the AI application according to whether the first device has the ability to The collaboration of the AI application can be performed. If so, the data of the input part of the model is transmitted to the first device through the data transmission channel of the transmission interface, so that the first device loads the input part of the AI algorithm model.
- the loading failure There may be various reasons for the loading failure.
- One possible reason is that, for example, the input part of the algorithm model corresponding to the AI application is not stored on the first device.
- Another possible reason is that the version of the input part of the algorithm model corresponding to the AI application stored on the first device is not the version required by the AI application.
- the model identifier of the algorithm model corresponding to the AI application can be used.
- the version identifier of the input part of the algorithm model corresponding to the AI application obtained from the device that stores the input part of the algorithm model corresponding to the AI application. Combining the above example, it may be obtained from the second device or obtained from a corresponding server, which is not limited here.
- the server takes the server in the network storing the input part and the output part of the algorithm model as an example.
- the server confirms whether the first device fails to load the input part of the algorithm model corresponding to the AI application
- the server confirms whether the first device can load the input part of the algorithm model corresponding to the AI application according to whether the first device has the ability to load the algorithm model. Execute the collaboration of the AI application. If so, the data of the input part of the model is transmitted to the first device through the data transmission channel of the transmission interface.
- the server when the server confirms that the second device fails to load the output part of the algorithm model corresponding to the AI application, the server confirms whether the second device can load the output part of the algorithm model corresponding to the AI application according to whether the second device has the ability to load and process the AI application. Execute the collaboration of the AI application. If so, the data of the output part of the model is transmitted to the second device through the data transmission channel of the transmission interface, so that the second device loads the output part of the algorithm model.
- the first feature extraction model in the multiple first devices is used as the input part of the algorithm model of the second device to realize the algorithm models of the multiple first devices and the second device.
- collaborative AI applications may also be the same feature extraction model, or may be different feature extraction models, and may be set accordingly based on specific applications, which is not limited here.
- the label information may be label information that is manually labeled through a network or human-computer interaction, or may be standard label information, clustering information, and other label information. Therefore, the AI algorithm model is trained by using the determined label data through the neural network training unit in the second device.
- the training process of the AI algorithm model can be performed offline or online on the server. It can also be trained or optimized online on the device or in the cloud.
- the parameters of the AI algorithm model can be updated through methods such as online learning, transfer learning, reinforcement learning, and federated learning to achieve model retraining or reoptimization.
- the model may also be obtained after collaborative training or optimization using the distributed system in this application. In this application, a specified model ID can be assigned to each AI algorithm model obtained after training.
- the distributed system of the present application is used to perform distributed collaborative training of AI algorithm models.
- the input information of the model is collected or input by the first device, and the transmission interface can transmit the NPU of the first device to output the training samples.
- the feature data for feature extraction uses the label information to train the feature data of the training samples sent by the first device to obtain the feedback data of the output part of the AI algorithm model, and the second device can use the AI algorithm model output part of the feedback
- the data is reversely transmitted to the input part of the model through the transmission interface, which is used to train the input part of the first device, so as to realize the coordinated adjustment of the network parameters of the two parts, so as to realize the coordinated training of the first device and the second device.
- the first device can perform feature extraction on the training samples according to the first feature extraction model to generate the first training feature data;
- the data processing apparatus sends the first training feature data.
- the second device trains the first feature data processing model according to the received first training feature data and the label information of the training samples, and determines feedback data for training the first feature extraction model. and send the feedback data to the first device.
- the first device trains the first feature extraction model according to the feedback data.
- an initial version ID can be assigned to the model obtained by offline training (for example, the initial version ID corresponds to the first feature extraction model and the first feature data processing model); if the model is retrained or reoptimized Then, update a specified version ID; after each model training or online optimization, assign an updated version ID to the input part and output part of the model (for example, the updated version ID corresponds to the second feature extraction model and The second feature data processing model).
- the input part and the output part of the model are collectively identified by the model ID and the version ID.
- corresponding flags can also be set. For example, for the input part of the model, the model ID and version ID and the input part ID can be set.
- the model ID and version ID and output part ID can be set.
- the model ID and version ID can be set correspondingly for the input part and output part of the model.
- the model ID includes the model input part ID and the model output part ID
- the version ID includes the input part ID of the model version and the output part ID of the model version. The specific implementation manner is not limited in this application.
- the first device executes the AI application collaboration corresponding to the AI algorithm model, it can perform feature extraction on the collected second media information based on the updated version, that is, according to the second feature extraction model, to obtain second feature data.
- the second characteristic data is sent to the second device.
- the second device can determine the second feature data processing model according to the model ID and the updated version ID, so as to obtain the reasoning result of the AI algorithm model of the AI application according to the second feature data processing model and the second feature data .
- the second device may receive training data of multiple first devices to train the neural network model of the second device, and correspondingly generate models in multiple first devices training feedback data, and send the plurality of feedback data to a plurality of first devices respectively, so that the plurality of first devices can use the corresponding feedback data to perform model training.
- unit identifiers may be set for each unit. Therefore, the second device can address and send control information and feedback data for training the AI algorithm model according to the unit ID of each unit.
- the sent control information message may carry the unit ID of the media information collection unit.
- the sent control information message may carry the unit ID of the neural network training unit.
- the first device 200 may include a processor 210 , an external memory interface 220 , an internal memory 221 , a transmission interface 230 , a media information collection unit 240 , and a communication unit 250 .
- the media information collection unit 240 may include: a microphone, an earphone interface, an audio unit, a speaker; a sensor unit 280, a camera 281, and the like.
- the communication unit 250 may include: antenna 1 , antenna 2 , a mobile communication unit 251 , and a wireless communication unit 252 .
- the processor 210 may include one or more processing units, for example, the processor 210 may include an application processor (application processor, AP), a modem processor, a graphics processor (graphics processing unit, GPU), an image signal processor (image signal processor, ISP), controller, memory, video codec, digital signal processor (digital signal processor, DSP), baseband processor, and neural network processing unit, such as neural network processor (Neural-network Processing) Unit, NPU), etc.
- different processing units may be independent devices, or may be integrated in one or more processors.
- the controller may be the nerve center and command center of the first device 200. The controller can generate an operation control signal according to the instruction operation code and timing signal, and complete the control of fetching and executing instructions.
- the neural network processor may include a neural network processing unit.
- the neural network processing unit is used to load the input part of the AI algorithm model corresponding to the first device, process the media information input to the neural network processing unit (for example, the media information obtained by the first device), and output the characteristics of the media information data.
- the first device uses the input part of the AI algorithm model loaded on the neural network processing unit of the first device to process the media information, and obtain abstract feature data after processing; the first device transmits the feature information to the second device through the transmission interface of the first device , so that the second device uses the output part of the model for further processing and obtains the output result of the AI application.
- the AI algorithm model may be determined by the server training, may also be determined by the independent training of the first device or the second device, or may be determined by the collaborative training of the first device and the second device.
- AI algorithm models can be trained offline or online. Based on the input portion of the AI algorithm model used by the first device, the output portion of the AI algorithm model is used by the second device. Therefore, correspondingly, the training process may also be the input part of the first device training the AI algorithm model, and the second device training the output part of the AI algorithm model, so as to realize the collaborative training of the AI algorithm model.
- the neural network processor may further include: a neural network training unit.
- the neural network training unit enables the first device to use the labeled data to train the AI algorithm model.
- the training process may be performed offline on the first device, may also be performed online at the end of the first device, or may be performed in coordination with the second device.
- the training data obtained from online training can be sent to the second device, so that the second device can train the model of the second device according to the training data obtained by the first device, and generate a model for the first device.
- Feedback data for model training so that the first device can train the neural network model of the first device according to the feedback data.
- a memory may also be provided in the processor 210 for storing instructions and data.
- the memory in processor 210 is cache memory.
- the memory may hold instructions or data that have just been used or recycled by the processor 210 . If the processor 210 needs to use the instruction or data again, it can be called directly from the memory. Repeated accesses are avoided, and the waiting time of the processor 210 is reduced, thereby improving the efficiency of the system.
- the processor 210 may execute the media information transmission method provided by the embodiment of the present application, so as to realize the collaboration of multiple devices of the first device under the AI application, and improve the user experience. After the processor 210 executes the media information transmission method provided by the embodiment of the present application, the processor 210 may generate and send feature data according to the acquired media information. It can also receive a control instruction from the second device to control the media acquisition unit of the first device. Optionally, when the first device includes a display screen, it can also receive media information from the second device and play the media. When the information is instructed, the media information can be played through the display screen.
- the processor 210 may include different devices.
- the CPU and the GPU may cooperate to execute the media information transmission method provided by the embodiments of the present application.
- some algorithms in the media information transmission method are executed by the CPU, and another part of the algorithms are executed by the GPU. Execute for faster processing efficiency.
- Internal memory 221 may be used to store computer executable program code, which includes instructions.
- the processor 210 executes various functional applications and data processing of the first device 200 by executing the instructions stored in the internal memory 221 .
- the internal memory 221 may include a storage program area and a storage data area.
- the storage program area may store operating system, code of application programs (such as camera application, WeChat application, etc.), and the like.
- the storage data area may store data created during the use of the first device 200 (such as images, videos, etc. captured by a camera application) and the like.
- the internal memory 221 may also store one or more computer programs corresponding to the data transmission algorithms provided in the embodiments of the present application.
- the one or more computer programs are stored in the aforementioned memory 221 and configured to be executed by the one or more processors 210, and the one or more computer programs include instructions that can be used to perform the execution of Figs. 7b For each step in the corresponding embodiment, the computer program can be used to implement the media information transmission method involved in the embodiment of the present application.
- the processor 210 may execute the media information transmission method involved in the embodiments of the present application.
- the internal memory 221 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, universal flash storage (UFS), and the like.
- non-volatile memory such as at least one magnetic disk storage device, flash memory device, universal flash storage (UFS), and the like.
- the code of the data transmission algorithm provided in the embodiment of the present application may also be stored in an external memory.
- the processor 210 may run the code of the data transmission algorithm stored in the external memory through the external memory interface 220, and the processor 210 may execute the media information transmission method involved in the embodiments of the present application.
- Camera 281 (front camera or rear camera, or a camera that can be both a front camera and a rear camera) is used to capture still images or video.
- the camera 281 may include a photosensitive element such as a lens group and an image sensor, wherein the lens group includes a plurality of lenses (convex or concave) for collecting light signals reflected by the object to be photographed, and transmitting the collected light signals to the image sensor .
- the image sensor generates an original image of the object to be photographed according to the light signal.
- the first device 200 may implement audio functions through an audio unit 270, a speaker 270A, a receiver 270B, a microphone 270C, an earphone interface 270D, an application processor, and the like. Such as music playback, recording, etc.
- the first device 200 may receive a key 290 input, and generate a key signal input related to user settings and function control of the first device 200 .
- the sensor unit 280 may include a distance sensor, a gyroscope sensor, an acceleration sensor, a proximity light sensor, a fingerprint sensor, a touch sensor, a temperature sensor, a pressure sensor, a distance sensor, a magnetic sensor, an ambient light sensor, an air pressure sensor, a bone conduction sensor, and the like, Not shown in the figure.
- the transmission interface in the first device 200 is used to connect other devices, so that the first device 200 and other devices transmit media information.
- the transmission interface in the first device 200 may include a first transmission interface and a third transmission interface.
- the data to be sent to the second device 300 is encapsulated into first bit stream data by being connected to the processor of the first device through the first transmission interface, and sent to the third transmission interface of the second device through the third transmission interface.
- the second bit stream data sent from the second device can be received through the third transmission interface, so that the data or message corresponding to the second bit stream data can be obtained through decapsulation through the first transmission interface (the second bit stream data is the second device data or messages encapsulated via the second transport interface). Therefore, the transmission channel established through the third transmission interface of the first device and the third transmission interface of the second device can support bidirectional transmission.
- the multiple first devices 200 may further encapsulate the first feature data sent to the second device through the first transmission interface of the multiple first devices.
- M pieces of first feature data may be encapsulated into independent bit stream data (which may be M pieces of first feature data) through the first transmission interface in the N first devices.
- the first characteristic data is packaged into M bit stream data, or it can be packaged into N bit stream data respectively according to N first devices), and the encapsulated bit stream data is packaged into a data through the third transmission interface.
- the packet (eg, the fourth message) is sent to the third transport interface of the second device.
- the second device can receive the fourth messages of the M first characteristic data of the N first devices that are encapsulated respectively through the third transmission interface, and decapsulate the fourth messages through the second transmission interface to obtain the M first characteristic data. feature data, and forward the M pieces of first feature data to the corresponding feature data processing model for processing according to the feature data processing model corresponding to the M pieces of first feature data to obtain the result of the first application.
- a cable suitable for the transmission interface 230 can be connected to and separated from the first device 200 by being inserted into the transmission interface 230 or unplugged from the transmission interface 230 .
- the wireless communication of the first device 200 can be realized through the antenna 1, the antenna 2, the mobile communication unit 251, the wireless communication unit 252, the modem processor and the baseband processor, etc. Function.
- Antenna 1 and Antenna 2 are used to transmit and receive electromagnetic wave signals.
- Each antenna in the first device 200 may be used to cover a single or multiple communication frequency bands. Different antennas can also be reused to improve antenna utilization.
- the antenna 1 can be multiplexed as a diversity antenna of the wireless local area network. In other embodiments, the antenna may be used in conjunction with a tuning switch.
- the mobile communication unit 251 may provide a wireless communication solution including 2G/3G/4G/5G etc. applied on the first device 200 .
- the mobile communication unit 251 may include at least one filter, switch, power amplifier, low noise amplifier (LNA), and the like.
- the mobile communication unit 251 can receive electromagnetic waves from the antenna 1, filter and amplify the received electromagnetic waves, and transmit them to the modulation and demodulation processor for demodulation.
- the mobile communication unit 251 can also amplify the signal modulated by the modulation and demodulation processor, and then turn it into an electromagnetic wave for radiation through the antenna 1 .
- at least part of the functional units of the mobile communication unit 251 may be provided in the processor 210 .
- the mobile communication unit 251 may be provided in the same device as at least part of the units of the processor 210 .
- the mobile communication unit 251 may also be used to perform information interaction with the second device, that is, send a media information transmission request to the second device, and encapsulate the sent media information transmission request into a message in a specified format , or the mobile communication unit 251 may be configured to receive a media information transmission request sent by the second device.
- the modem processor may include a modulator and a demodulator.
- the modulator is used to modulate the low frequency baseband signal to be sent into a medium and high frequency signal.
- the demodulator is used to demodulate the received electromagnetic wave signal into a low frequency baseband signal. Then the demodulator transmits the demodulated low-frequency baseband signal to the baseband processor for processing.
- the low frequency baseband signal is processed by the baseband processor and passed to the application processor.
- the modulation and demodulation processor may further include a channel coding unit and a decoding unit, wherein the channel coding unit is configured to perform channel coding on the data signal acquired by the first device according to the link layer and physical layer protocols to obtain the physical layer signal, and convert the physical layer signal to the physical layer.
- the signal is transmitted to the transmission interface.
- the data signal may be feature data determined by performing feature extraction on media information, or may be data of media information.
- a control information decoding unit may also be included, configured to decode the control signal sent by the second device, and may also be configured to decode received feedback data from the second device, where the feedback information is used for online training and optimization of the model.
- the application processor outputs sound signals through audio devices (not limited to speakers 270A, receivers 270B, etc.), or displays images or videos through a display screen.
- the modem processor may be a stand-alone device. In other embodiments, the modem processor may be independent of the processor 210, and may be provided in the same device as the mobile communication unit 251 or other functional units.
- the wireless communication unit 252 may provide applications on the first device 200 including wireless local area networks (WLAN) (such as wireless fidelity (Wi-Fi) networks), bluetooth (BT), global navigation Satellite system (global navigation satellite system, GNSS), frequency modulation (frequency modulation, FM), near field communication technology (near field communication, NFC), infrared technology (infrared, IR) and other wireless communication solutions.
- WLAN wireless local area networks
- BT wireless fidelity
- GNSS global navigation Satellite system
- frequency modulation frequency modulation, FM
- NFC near field communication technology
- IR infrared technology
- the wireless communication unit 252 may be one or more devices integrating at least one communication processing unit.
- the wireless communication unit 252 receives electromagnetic waves via the antenna 2 , frequency modulates and filters the electromagnetic wave signals, and sends the processed signals to the processor 210 .
- the wireless communication unit 252 can also receive the signal to be sent from the processor 210 , perform frequency modulation on it, amplify it, and convert it into an electromagnetic wave for radiation through the antenna 2 .
- the wireless communication unit 252 is configured to establish a connection with the second device, and cooperate to complete the task of the AI application through the second device.
- the wireless communication unit 252 may be configured to access the access point device, send a message corresponding to a request for transmission of media information of feature data to the second device, or receive a message corresponding to a request for transmission of media information sent from the second device.
- the wireless communication unit 252 can also be used to receive media information from other devices.
- FIG. 3b it is a schematic structural diagram of a second device according to an embodiment of the present application.
- the second device 300 may include a processor 310 , an external memory interface 320 , an internal memory 321 , a transmission interface 330 , an antenna 11 , an antenna 12 , a mobile communication unit 351 , and a wireless communication unit 352 .
- the processor 310 may include one or more processing units, for example, the processor 310 may include an application processor (application processor, AP), a modem processor, a graphics processor (graphics processing unit, GPU), an image signal processor (image signal processor, ISP), controller, memory, video codec, digital signal processor (digital signal processor, DSP), baseband processor, and neural network processor (Neural-network Processing Unit, NPU), etc. Wherein, different processing units may be independent devices, or may be integrated in one or more processors.
- the controller may be the nerve center and command center of the second device 300 . The controller can generate an operation control signal according to the instruction operation code and timing signal, and complete the control of fetching and executing instructions.
- the neural network processor may include a neural network processing unit.
- the neural network processing unit is used to load the output part of one or more AI algorithm models. According to the output part of the AI algorithm model corresponding to the feature data, process the feature data, transmit the decoded feature data to the neural network processing unit of the second device, and use the output part of the loaded AI algorithm model of the neural network unit to process the feature data.
- the data is processed to obtain the final inference results processed by the AI algorithm; the inference results such as detection, classification, identification, positioning, and tracking are obtained.
- the inference results are output to the artificial intelligence application, and the artificial intelligence application uses the inference results to realize functions such as biometric recognition, environment recognition, scene modeling, machine vision interaction, and voice interaction.
- the second device in the embodiment of the present application may further include: a neural network training unit, where the neural network training unit enables the second device to train the AI algorithm model by using the labeled data.
- the training process may be performed offline on the second device, or performed online at the second device, and may also be performed in collaboration with the first device.
- the feedback data obtained from the online training can be sent from the model output part to the model input part through the transmission interface through the interface system, So that the first device trains the neural network model of the first device according to the feedback data.
- the second device may also send the online training feedback data of the AI model to the first device.
- the second device may receive training data of multiple first devices to train the neural network model of the second device, and correspondingly generate models in multiple first devices training feedback data, and send the plurality of feedback data to a plurality of first devices respectively, so that the plurality of first devices can use the corresponding feedback data to perform model training.
- a memory may also be provided in the processor 310 for storing instructions and data.
- the memory in processor 310 is cache memory. This memory may hold instructions or data that have just been used or recycled by the processor 310 . If the processor 310 needs to use the instruction or data again, it can be called directly from the memory. Repeated accesses are avoided, and the waiting time of the processor 310 is reduced, thereby increasing the efficiency of the system.
- the processor 310 may execute the media information transmission method provided by the embodiment of the present application, so as to realize the collaboration between the second device and the first device under the AI application, and improve the user experience. After the processor 310 executes the media information transmission method provided by the embodiment of the present application, the processor 310 may generate and send feature data according to the acquired media information. A control instruction may also be sent to the first device for controlling the media acquisition unit of the first device. Optionally, when the first device includes a display screen, media information and a message for playing the media information may also be sent to the first device. When instructed, the media information can be played through the display screen.
- the second device when the second device includes a display screen, it can also receive media information sent by the first device, or after receiving the feature data from the first device, process the feature data, and obtain the inference processed by the AI algorithm.
- the media information can be played through the display screen.
- the processor 310 may include different devices. For example, when a CPU and a GPU are integrated, the CPU and the GPU may cooperate to execute the media information transmission method provided by the embodiments of the present application. For example, some algorithms in the media information transmission method are executed by the CPU, and another part of the algorithms are executed by the GPU. Execute for faster processing efficiency.
- Internal memory 321 may be used to store computer executable program code, which includes instructions.
- the processor 310 executes various functional applications and data processing of the second device 300 by executing the instructions stored in the internal memory 321 .
- the internal memory 321 may include a storage program area and a storage data area.
- the storage program area may store operating system, code of application programs (such as camera application, WeChat application, etc.), and the like.
- the storage data area may store data created during the use of the second device 300 (such as images, videos, etc. captured by a camera application) and the like.
- the internal memory 321 may also store one or more computer programs corresponding to the data transmission algorithms provided in the embodiments of the present application.
- the one or more computer programs are stored in the aforementioned memory 321 and configured to be executed by the one or more processors 310, and the one or more computer programs include instructions that can be used to perform the execution of Figs. 2a-2a 7a
- the computer program can be used to implement the media information transmission method involved in the embodiment of the present application.
- the processor 310 may execute the media information transmission method involved in the embodiments of the present application.
- the internal memory 321 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, universal flash storage (UFS), and the like.
- non-volatile memory such as at least one magnetic disk storage device, flash memory device, universal flash storage (UFS), and the like.
- the code of the data transmission algorithm provided in the embodiment of the present application may also be stored in an external memory.
- the processor 310 may run the code of the data transmission algorithm stored in the external memory through the external memory interface 320, and the processor 310 may execute the media information transmission method involved in the embodiments of the present application.
- the transmission interface 330 in the second device 300 is used to connect other devices, so that the second device 300 and other devices transmit media information.
- the transmission interface in the second device 300 may include a second transmission interface and a third transmission interface.
- the second transmission interface is connected to the processor of the second device, and the data to be sent to the second device 300 is encapsulated into second bit stream data through the second transmission interface, and sent to the third transmission interface of the first device through the third transmission interface. transport interface.
- the first bit stream data from the first device can be received through the third transmission interface, and the feature data, data sum, control information, feedback data, handshake data, messages, etc. sent by the first device can be obtained through decapsulation through the second transmission interface . Therefore, the transmission channel established through the third transmission interface of the first device and the third transmission interface of the second device can support bidirectional transmission.
- the second device 300 may also receive a fourth message through the second transmission interface of the second device; the fourth message includes M pieces of first feature data of the N first devices; N, M is a positive integer greater than 1; M is greater than or equal to N; specifically, the fourth message of the M first feature data of the N first devices encapsulated respectively can be received through the third transmission interface (the fourth message may be is encapsulated into N data packets, or can be encapsulated into M data packets, which is not limited here), and decapsulates the fourth message through the second transmission interface to obtain M first feature data, and according to the M A feature data processing model corresponding to the first feature data, the M pieces of first feature data are forwarded to the corresponding feature data processing model for processing, and a result of the first application is obtained.
- the fourth message includes M pieces of first feature data of the N first devices; N, M is a positive integer greater than 1; M is greater than or equal to N; specifically, the fourth message of the M first feature data of the N first devices encapsulated respectively can be
- the transmission interface 330 is a wired transmission interface
- a cable suitable for the transmission interface can be inserted into the transmission interface 330 or pulled out from the transmission interface 330 to achieve contact with and separation from the second device 300 .
- the wireless communication function of the second device 300 can use the antenna 11, the antenna 12, the mobile communication unit 351, the wireless communication unit 352, the modem processor and the baseband processor, etc. accomplish.
- the antenna 11 and the antenna 12 are used to transmit and receive electromagnetic wave signals.
- Each antenna in the second device 300 may be used to cover a single or multiple communication frequency bands. Different antennas can also be reused to improve antenna utilization.
- the antenna 11 can be multiplexed into a diversity antenna of the wireless local area network.
- the antenna may be used in conjunction with a tuning switch.
- the mobile communication unit 351 may provide a wireless communication solution including 2G/3G/4G/5G etc. applied on the second device 300 .
- the mobile communication unit 351 may include at least one filter, switch, power amplifier, low noise amplifier (LNA), and the like.
- the mobile communication unit 351 can receive electromagnetic waves from the antenna 11, filter, amplify, etc. the received electromagnetic waves, and transmit them to the modulation and demodulation processor for demodulation.
- the mobile communication unit 351 can also amplify the signal modulated by the modulation and demodulation processor, and then convert it into electromagnetic waves and radiate it out through the antenna 11 .
- at least part of the functional units of the mobile communication unit 351 may be provided in the processor 310 .
- the mobile communication unit 351 may be provided in the same device as at least part of the units of the processor 310 .
- the mobile communication unit 351 may also be used to perform information interaction with the second device, that is, to send a media information transmission request to the first device, and the received media information transmission request may be encapsulated into a message in a specified format , or the mobile communication unit 351 may be used to transmit a media information transmission instruction to the first device or a control information message to the first device.
- the modem processor may include a modulator and a demodulator.
- the modulator is used to modulate the low frequency baseband signal to be sent into a medium and high frequency signal.
- the demodulator is used to demodulate the received electromagnetic wave signal into a low frequency baseband signal. Then the demodulator transmits the demodulated low-frequency baseband signal to the baseband processor for processing.
- the low frequency baseband signal is processed by the baseband processor and passed to the application processor.
- the modem processor may also include a channel encoding unit and a decoding unit.
- the channel decoding unit can decode the data signal sent by the first device from the received physical layer signal of the first device according to the link layer and the physical layer protocol.
- the data signal may be feature data determined by the first device by performing feature extraction on the media information (the feature data may be used as an input of the neural network processing unit), or may be data of media information.
- the feature data may be used as an input of the neural network processing unit
- the second device can also be transmitted through a corresponding transmission channel through a transmission interface according to a corresponding transmission protocol. These information may be sent together with the media information, or may be sent through other transmission protocols, and the specific implementation manner is not limited in this application.
- the channel encoding unit may be used to encode the data signal sent by the second device.
- the data signal may be a control command sent to the first device.
- the control command can be channel-coded according to the interface transmission protocol through the channel coding unit of the second device; the encoded control signal is modulated by the transmission interface and sent to the control channel, and transmitted to the first device through the transmission interface and the control channel of the second device , so that the first device can receive the control instruction through the control channel.
- the channel coding unit may also be used for coding feedback data sent to the first device, where the feedback information is used for online training and optimization of the model of the first device.
- the application processor outputs a sound signal through an audio device, or displays an image or video through a display screen.
- the modem processor may be a stand-alone device. In other embodiments, the modem processor may be independent of the processor 310, and may be provided in the same device as the mobile communication unit 351 or other functional units.
- the wireless communication unit 352 can provide applications on the second device 300 including wireless local area networks (WLAN) (such as wireless fidelity (Wi-Fi) networks), bluetooth (BT), global navigation Satellite system (global navigation satellite system, GNSS), frequency modulation (frequency modulation, FM), near field communication technology (near field communication, NFC), infrared technology (infrared, IR) and other wireless communication solutions.
- WLAN wireless local area networks
- BT wireless fidelity
- GNSS global navigation Satellite system
- frequency modulation frequency modulation, FM
- NFC near field communication technology
- infrared technology infrared, IR
- the wireless communication unit 352 may be one or more devices integrating at least one communication processing unit.
- the wireless communication unit 352 receives electromagnetic waves via the antenna 12 , frequency modulates and filters the electromagnetic wave signals, and sends the processed signals to the processor 310 .
- the wireless communication unit 352 can also receive the signal to be sent from the processor 310 , perform frequency modulation on it, amplify it, and convert it into electromagnetic waves for radiation through the antenna 12 .
- the wireless communication unit 352 is configured to establish a connection with the first device, and complete the task of the AI application by cooperating with the first device.
- the wireless communication unit 352 may also be configured to access the access point device, receive a message corresponding to a transmission request for feature data sent by the first device, or send a message corresponding to a request for transmission of media information to the first device , and send a message corresponding to the control information to the first device.
- the wireless communication unit 352 may also be configured to receive media information from other first devices or information of other devices.
- FIG. 3c it is a schematic diagram of the architecture of a distributed system for AI application collaboration composed of a first device and a second device according to an embodiment of the present application.
- the transmission interfaces provided in the embodiments of the present application may be the transmission interface 230 in the first device or the communication unit 250 of the first device, and the transmission interface 330 in the second device or the communication unit 350 of the second device).
- the transmission interface 230 in the first device and the transmission interface 330 in the second device are taken as examples for description.
- the transmission interface provided in this embodiment of the present application may be applicable to multiple transmission protocols, and may also be called an aggregation interface, or may be called a new interface (NEW interface), or may use other names, which are not limited in this embodiment of the present application.
- NGW interface new interface
- the transmission interface protocol in this embodiment of the present application supports the transmission of feature data output by the first device.
- the feature data obtained after the input part of the AI model is processed is much lower than the original audio and video media data, so it occupies a lower bandwidth, which solves the problem of high bandwidth transmission and consumption of more bandwidth resources.
- real-time transmission of low-bandwidth transmission interfaces is realized, creating conditions for the realization of real-time distributed AI processing.
- the original media information cannot be recovered through the transmitted characteristic data, the potential risk of privacy data leakage can be solved, and the data security of the media information transmission can be improved.
- the transmission interface protocol in the embodiments of the present application may support the first device and the second device to perform software processing of AI processing such as the NPU architecture, the model identifier (ID) and version identifier (ID) of the loaded AI algorithm model, etc.
- the hardware capability is negotiated to determine whether it is possible to coordinate and form a distributed system for AI application collaboration to complete the processing tasks of the corresponding AI application; during the capability negotiation process, the transmission interface supports bidirectional transmission of AI algorithm models. Part or all of the AI algorithm model stored by the device or the second device or the AI algorithm model obtained from the network is transmitted to the first device or the second device, so as to realize the loading of the input part and the output part of the AI algorithm model.
- the online training feedback data output by the second device can be sent to the first device for online training or online training of the input information of the AI algorithm model on the first device. Online optimization, and feedback the feature data of further training to the second device, so as to realize online training and online optimization of the AI algorithm model of the first device and the second device.
- the transmission interface may also be used to transmit multiple types of data.
- the transmission interface can also transmit the output feature data of the input part of the AI model, and can also transmit control messages such as handshake signals and control signals in both directions.
- the transmission interface may also transmit media information signals or other data signals.
- the transmission interface can perform simultaneous aggregate transmission and bidirectional transmission of the above-mentioned signals.
- the transmission interface can support compatible transmission of media information and AI feature data, the transmission interface can transmit standard media information data, and also has the ability to transmit feature data processed by the NPU at the acquisition end.
- the transmission interface can also transmit media signals and other data signals, and the transmission interface can also transmit handshake and control signals.
- the transmission interface may be a unidirectional transmission interface or a bidirectional transmission interface. Taking a unidirectional transmission interface as an example, a sending interface is set at the sending end, and a receiving interface is set at the receiving end. Thus, the function of transmitting media data from the sender to the receiver is realized.
- the transmission interface may be a bidirectional transmission interface.
- the transmission interface has a sending function and a receiving function, that is, supports bidirectional data transmission.
- the transmission interface supports sending and receiving data signals, that is, the transmission interface can be used as both a sending end of a data signal and a receiving end of the data signal.
- the transmission interface has the capability of data aggregation transmission.
- the protocol of the interface can support the simultaneous transmission of media information and AI feature data in the same channel by technologies such as data packaging and mixing.
- the transport interface may transmit raw or compressed media information, for example, the transport interface may support bidirectional transmission of media information and AI feature data by configuring multiple channels.
- the transmission interface in the first device 200 may include a first transmission interface and a third transmission interface.
- the data to be sent to the second device 300 is encapsulated into first bit stream data by being connected to the processor of the first device through the first transmission interface, and sent to the third transmission interface of the second device through the third transmission interface.
- the second bit stream data sent from the second device can be received through the third transmission interface, so that the data or message corresponding to the second bit stream data can be obtained through decapsulation through the first transmission interface (the second bit stream data is the second device data or messages encapsulated via the second transport interface).
- the transmission interface in the second device 300 may include a second transmission interface and a third transmission interface.
- the second transmission interface is connected to the processor of the second device, and the data to be sent to the second device 300 is encapsulated into second bit stream data through the second transmission interface, and sent to the third transmission interface of the first device through the third transmission interface. transport interface.
- the first bit stream data from the first device can be received through the third transmission interface, and the feature data, data sum, control information, feedback data, handshake data, messages, etc. sent by the first device can be obtained through decapsulation through the second transmission interface . Therefore, the transmission channel established through the third transmission interface of the first device and the third transmission interface of the second device can support bidirectional transmission.
- the third transmission interface in the plurality of first devices 200 is a third transmission interface, and in this case, the plurality of first bit streams encapsulated by the first transmission interfaces in the plurality of first devices
- the data may be uniformly sent to the third transmission interface of the second device 300 through the third transmission interface.
- N first devices generate M pieces of first feature data, and respectively encapsulate the M pieces of first feature data through N first transmission interfaces of the N first devices, and package them into a fourth message through a third transmission interface .
- the fourth message is received through the third transmission interface of the second device 300, so that the M pieces of first characteristic data are decapsulated through the second transmission interface, and according to the characteristic data processing model corresponding to the M pieces of first characteristic data, the The M pieces of first feature data are forwarded to the corresponding feature data processing model for processing.
- the data signal here may be multimedia data, may also be feature data involved in the embodiments of the application, may also be control information used to establish a transmission link, or may be used to transmit other parameters and Other data signals are not limited here.
- the following example illustrates the specific transmission process of the transmission interface in this application.
- the transmitting end compresses and encrypts the transmitted data; transmits the transmitted data to the transmission interface through channel coding, and then transmits the transmitted data to the physical layer channel of the interface after modulation.
- the first device may further include a channel coding unit, which performs channel coding on the characteristic data according to a transmission interface or a data transmission protocol agreed in a standard by the channel coding unit to obtain a coded signal.
- the first device may further compress the feature data to further reduce the amount of data transmission.
- the first device when it performs channel coding on the feature data, it can also encrypt the channel-coded feature data, and modulate the channel-coded feature data according to the transmission interface system or the electrical layer and physical layer transmission protocols agreed by the standard. The transport channel to which the physical layer signal is sent from the output interface.
- the receiving end after the interface of the receiving end demodulates the physical layer signal, it can perform channel decoding to obtain the transmitted data; correspondingly, the receiving end can also decompress and decrypt the decoded signal.
- the second device may further include a channel decoding unit for receiving characteristic data of the first device.
- the physical layer signal may be received from the transmission channel through the input interface of the second device, and the encoded signal may be obtained by demodulating the physical layer signal.
- the channel decoding unit performs channel decoding on the received coded signal according to the protocol of the transmission interface, and obtains characteristic data sent by the second device.
- the channel decoding unit may further decrypt and decompress the encoded signal.
- the first device 200 or the second device 300 may further include: a display screen for displaying images, videos, and the like.
- the display includes a display panel.
- the display panel can be a liquid crystal display (LCD), an organic light-emitting diode (OLED), an active-matrix organic light-emitting diode or an active-matrix organic light-emitting diode (active-matrix organic light).
- LED liquid crystal display
- OLED organic light-emitting diode
- AMOLED organic light-emitting diode
- FLED flexible light-emitting diode
- Miniled MicroLed, Micro-oLed
- quantum dot light-emitting diode quantum dot light emitting diodes, QLED
- the first device 200 or the second device 300 may include 1 or S display screens, where S is a positive integer greater than 1.
- the display screen can be used to display information input by the user or information provided to the user (for example, video information, voice information, image information, text information, etc.) and various graphical user interfaces (graphical user interface, GUI).
- GUI graphical user interface
- the display screen can display photos, videos, web pages, or documents, and the like.
- the display screen may display a graphical user interface as shown in Figure 4b.
- the graphical user interface shown in FIG. 4b may include a status bar, a hideable navigation bar, a time and weather widget (widget), and an application icon, such as a browser icon.
- the status bar includes operator name (eg China Mobile), mobile network (eg 4G), time and remaining battery.
- the navigation bar includes a back button icon, a home button icon, and a forward button icon.
- the status bar may further include a Bluetooth icon, a Wi-Fi icon, an external device icon, and the like.
- the graphical user interface shown in FIG. 4b may further include a Dock bar, and the Dock bar may include commonly used application icons and the like.
- the processor 210 After the processor 210 detects a touch or gesture event of a user's finger (or a stylus, etc.) on an application icon, in response to the touch or gesture event, the user interface of the application corresponding to the application icon is opened, and the The user interface of the application is shown on the display.
- the display screen of the first device 200 or the second device 300 displays a main interface, and the main interface includes icons of multiple applications (such as a camera application, a WeChat application, etc.).
- the monitor displays the interface of camera applications, such as the viewfinder interface.
- the first device 200 or the second device 300 may use a motor to generate a vibration prompt (eg, a vibration prompt for an incoming call).
- the indicator in the first device 200 or the second device 300 may be an indicator light, which may be used to indicate a charging state, a change in power, or a message, a missed call, a notification, and the like.
- the first device 200 or the second device 300 may implement an audio function through an audio unit, an application processor, and the like. Such as music playback, recording, etc.
- the audio unit may include one or more of a speaker, a receiver, a microphone, and an earphone jack.
- the first device 200 or the second device 300 may receive a key input, and generate a key signal input related to user settings and function control of the first device 200 or the second device 300 .
- the display screen may be an integrated flexible display screen, or a spliced display screen composed of two rigid screens and a flexible screen located between the two rigid screens may be used.
- the first device 200 may include more or less components than those shown in FIG. 3a, and the second device 300 may include more or less components than those shown in FIG. 3b.
- the embodiment is not limited.
- the illustration of the first device 200 or the second device 300 is only an example, and the first device 200 or the second device 300 may have more or fewer components than those shown in the figures, two or more may be combined multiple components, or may have different component configurations.
- the various components shown in the figures may be implemented in hardware, software, or a combination of hardware and software, including one or more signal processing and/or application specific integrated circuits.
- FIG. 4a is a schematic flowchart of a method for transmitting media information according to an embodiment of the present application
- a first device and a second device establish a communication connection for AI application collaboration, and through the cooperation between the first device and the second device, The first device and the second device can cooperate to complete an AI task.
- the following is an example of a communication connection mode in which the second device has a display screen and the second device initiates the establishment of a distributed system for AI application collaboration.
- the first device may also actively initiate the establishment of a communication connection with the second device, which may be implemented with reference to the manner of this embodiment, which is not limited herein. Specifically, it can include:
- Step 401 The second device discovers the first device through the device discovery protocol.
- the first device and the second device can be connected to each other through their respective transmission interfaces and corresponding wired or wireless channels.
- a wireless communication connection can be established through Bluetooth, NFC, or WIFI, or a communication connection can be established through a wired method.
- Step 402 The second device sends a capability negotiation request message to the first device.
- the capability negotiation request message is used to request a transmission protocol supported by the first device, and the transmission protocol of the first device is used to indicate whether the first device supports transmission of feature data.
- the second device may discover the first device according to the device discovery protocol adapted in the embodiment of the present application. Further, the capability negotiation between the first device and the second device is carried out through the control channel and the handshake protocol in the transmission interface, so that the second device determines the type of the first device, the software capability that supports AI processing, the hardware capability that supports AI processing, and the Information such as the transmission protocol of the media information, so as to determine whether the first device can establish a communication connection with the second device, so as to realize the function of corresponding AI application collaboration.
- the establishment of a distributed system composed of the first device and the second device may be triggered by a corresponding AI application, or the establishment of a distributed system composed of the first device and the second device may be triggered by other methods, or It is a distributed system composed of a first device and a second device established by a user initiatively initiates the establishment of a communication connection between the first device and the second device.
- the user may correspondingly set the AI application coordination function of the first device or the second device through the first device or the second device provided with the interface of the AI application coordination function.
- an interface corresponding to the AI application coordination function of the first device may be set on the first device, so that the user can set the AI application coordination function of the first device on the interface.
- An interface corresponding to the AI application coordination function of the second device may also be set on the second device, so that the user can set the AI application coordination function of the first device on the interface.
- the AI application collaboration function of the first device and the second device can be set on the device that has the function of displaying the interface.
- the first device and the second device as a whole perform AI application collaboration function setting.
- the AI application coordination function of the first device can be set, and the AI application coordination function of the second device can also be set, so as to determine the composition of the first device and the second device.
- the AI application collaboration function is set, which is not limited here.
- the following takes an example of a manner in which the user actively triggers the process.
- the second device may be provided with an interface with an AI application collaboration function.
- an AI application collaboration function exemplary, as shown in FIG. 4b, it is the control interface 410 of the AI application coordination function of the second device, and the user can operate the control interface 410 to set whether to establish the AI application coordination function with the first device.
- the user can enable/disable the AI application coordination function based on the enable/disable control 420 of the AI application coordination function.
- the user turns on the AI application collaboration function of the second device.
- the identifier of the first device discovered by the second device may be displayed in the interface of the AI application cooperation function of the second device.
- the identifier of the first device may include a device icon of the first device and/or a discovery name of the first device when the first device acts as a feature data sender for AI application collaboration.
- the second device in response to a user's click operation, displays a first device search interface 430, which may be a device search interface of the AI application cooperative service of the second device, and the device search interface 430 may include the identification of the first device that can be discovered.
- the search device interface includes the icon 450 of the first device and the discovery of the first device as the feature data sender of AI application collaboration. Name 440.
- the first device may be a smart screen 1, a camera 1, an earphone 1, and AR glasses.
- the discovered devices may not be distinguished according to the types of the devices of the collaboration server in the first device search interface.
- the user may operate the notification bar or the prompt box in the task bar of the second device, and in response to the above operation, the second device opens the above-mentioned first search device interface.
- the user may operate a relevant icon in the notification bar or task bar of the second device, and in response to the above operation, the second device opens the above-mentioned first search device interface.
- the second device may send a capability negotiation request to the first device.
- the capability negotiation request may be a handshake request message sent to the first device through a control channel in a transmission interface of the second device and a corresponding handshake protocol.
- the capability negotiation request is used to request capability information such as the media information acquisition capability of the first device, the main parameters of the media information, the AI software and hardware processing capability, and the transmission protocol supported by the transmission interface.
- Step 403 The first device sends a capability negotiation response message to the second device.
- the capability negotiation response message may be used to confirm that the first device supports a transmission protocol for transmitting characteristic data.
- the capability negotiation response message may also be used to confirm capability information of the first device.
- the capability negotiation response message further includes at least one of the following: the feature extraction capability of the first device, and the version of the feature extraction model in the first device.
- the first device may return a capability negotiation response to the second device, where the capability negotiation response message may include: the media information acquisition capability of the first device , the main parameters of media information, AI software and hardware processing capabilities, transmission protocols supported by the transmission interface and other capability information.
- the first device may determine, according to the capability negotiation response message, whether the first device can form a distributed system through respective transmission interfaces for AI application collaboration (for example, the first application collaboration). That is, whether the transmission interface of the first device and the transmission interface of the second device can support the transmission of feature data of media information, and whether the first device has the AI processing capability for feature extraction of media information, and whether the first device supports the corresponding AI application The processing power of the input portion of the model. _
- a capability negotiation confirmation message may be sent to the first device, where the capability negotiation confirmation message may be used to prompt the first device that capability negotiation is successful and whether to enable the AI application collaboration function.
- a reminder message of the capability negotiation confirmation message may also be displayed on the second device, as shown in (a) in Figure 4d, to prompt the user whether to turn on the first device Synergy function with the AI application of the second device.
- a setting control may be provided for jumping to the interface of AI application collaboration, so that the user can create AI application collaboration with the first device.
- the second device may return the capability negotiation failure to the first device message, as shown in (b) in Figure 4d and (c) in Figure 4d, on the interface of the reminder message, a view detail control can be set to jump to the interface of AI application collaboration to An interface that allows users to view the failure result of the specific AI application collaboration capability negotiation.
- the device may determine whether to still transmit media information with the first device according to the negotiation failure result, so as to realize the AI application function of the second device.
- Another possible way, as shown in (b) of FIG. 4d on the interface of the reminder message, can also prompt the user whether to transmit media information with the first device, so as to realize the AI application function of the second device.
- the first device and the second device may negotiate the capability of transmitting media information.
- the first device can establish a communication link for transmitting media information with the second device, and the first device can convert the media information based on the acquired media information and media encoding methods supported by the first device and the second device.
- the second device decodes the received media information signal to obtain the corresponding media information, and processes the media information according to the needs of the AI application.
- the second device can display/play the media information.
- the second device may perform media encoding and channel encoding on the processing result of the media information to generate a signal of the media information that can be displayed/played by the first device, Therefore, the first device can receive the signal of the media information and display/play.
- Step 404 The second device sends an authentication request message to the first device.
- the authentication request message is used to request whether the first device establishes a trusted communication connection with the data processing apparatus, and the communication connection is used to confirm the authority of the data processing apparatus to control the first device.
- the second device in response to the user's operation of the AI application collaboration request on the first interface, may send a security authentication request message to the first device; the security authentication request message is used to request the second device to obtain the first
- the control authority of the unit used for the AI application collaboration function such as the media collection unit of a device.
- the authentication method corresponding to the security authentication request may include manual authorization, unified biometric authentication authorization, account authorization, cloud authorization, near field communication authorization, and the like. Take an example of how a user enters a username and password.
- the second device may display an authentication interface of the first device, where the authentication interface is used to prompt the user to input a user name and password for logging in and authenticating the first device.
- the second device may carry the user name and password input by the user in a security authentication request message and send it to the first device.
- the first device receives the security authentication request message of the second device from the transmission port, and verifies the validity of the user name and password carried in the security authentication request.
- the first device may send a security authentication response message to the second device, and the security authentication response message is used to notify the second device whether the second device can obtain the first device.
- the control authority of the unit used for the AI application collaboration function such as the media collection unit of a device.
- Step 405 The first device sends an authentication response message to the second device.
- the authentication response message is used to confirm whether the first device establishes a trusted communication connection with the data processing apparatus.
- Step 406 The second device may send an authentication success message to the first device.
- authentication success message is optional.
- the authentication success message includes: the device identifier corresponding to the first device, and the identifiers of the distributed systems where the first device and the second device are located.
- the second device may assign an identifier of the distributed system to the distributed system composed of the first device and the second device in which the AI application cooperates, that is, it can be considered that The first device and the second device form a super device or distributed system.
- a corresponding device identifier may also be assigned to the first device, and a corresponding device identifier may be assigned to the second device for data transmission between the distributed systems.
- the identifier of the distributed system can be used for the second device to establish a communication link with the first device. And establish the corresponding media data link, the characteristic data transmission link obtained by the model corresponding to the AI application, and the control link according to the requirements.
- the second device can receive the feature data sent by the first device through the feature data transmission link, and after processing the feature data, it can be used for AI applications.
- the second device may receive the media information sent by the first device through the media data link.
- the second device can also send a control instruction to the first device through the control link, and the control instruction can be used to instruct the control of the units (media information collection unit, storage unit, media information playback unit, etc.) in the first device, For example, control methods such as the start and end of media information collection by the media information collection unit, and the parameter adjustment control and operation of the media information collection unit.
- control methods such as the start and end of media information collection by the media information collection unit, and the parameter adjustment control and operation of the media information collection unit.
- the first device cannot obtain the control authority of the unit required for the AI application collaboration corresponding to the first device, and can display a notification message to the second device that the AI application collaboration between the second device and the first device fails to establish.
- the security authentication part fails, for example, the authentication of the media acquisition unit in the first device succeeds, and the authentication of the storage unit fails, at this time, a communication connection for AI application collaboration can be established with the second device for the successfully authenticated unit. Further, on the interface of the AI application collaboration of the second device, it can be displayed that the authentication of the media collection unit of the second device and the first device is successful, and a notification message that the AI application collaboration is successful is established. For another example, the authentication of the AI processing unit of the first device is successful. In this case, the second device may establish a communication connection for AI application collaboration with the AI processing unit of the first device.
- the second device may configure device identifiers for the AI processing units of the second device and the first device to form a distributed system in which corresponding AI applications are coordinated.
- the device is used for the first device and the second device to establish a communication link of characteristic data, and can also establish a communication link of media information.
- the first device can send feature data to the second device, so that the second device operates on the output part of the model of the AI application according to the feature data sent by the first device, and obtains the inference result of the AI application, thereby realizing the relationship between the first device and the second device.
- the AI application collaboration of the two devices are used for the first device and the second device to establish a communication link of characteristic data, and can also establish a communication link of media information.
- the first device can send feature data to the second device, so that the second device operates on the output part of the model of the AI application according to the feature data sent by the first device, and obtains the inference result of the AI application, thereby realizing the relationship between the first device and the second device.
- both the first device and the second device may be multiple, and one first device may initiate the establishment of a communication connection coordinated by AI applications, or multiple devices may jointly initiate a capability negotiation request and security Authentication request, at this time, the first device and the second device can be respectively confirmed by the server to establish the capability of establishing AI application collaboration, and the first device and the second device can be safely authenticated.
- the control authority of the corresponding first device may be configured for the second device, and the device identifier may be configured for the composed distributed system.
- first devices When there are multiple first devices or multiple units such as media information collection units, storage units, and transmission interfaces of the first devices used for AI application collaboration in a distributed system composed of AI application collaboration, they can be different units.
- a display interface after the second device and the first device successfully establish AI application collaboration in the display interface, a unit that can establish AI application collaboration with the second device can be displayed, and Unit ID, unit status (whether AI application collaboration is enabled).
- a unit for establishing AI application collaboration with the second device may also be actively set by the user, which is not limited by application.
- the feature data obtained from the media information collected by multiple first devices or multiple media information collection units can be distinguished from the respective feature data according to the respective unit identifiers after feature extraction is performed on the input parts of the respective models. Therefore, it is aggregated through the transmission interface and then sent to the second device uniformly.
- the second device can determine the respective characteristic data according to the respective unit identifiers, thus, perform corresponding processing, and send the corresponding unit identifiers to the corresponding device by carrying the respective unit identifiers.
- the device sends a control instruction to realize AI application collaboration of multiple devices, so that the second device can perform unified control and management of multiple first devices or multiple units.
- the second device may also display a notification message to notify the user that the connection between the second device and the first device has been successfully established, which may Establish a transmission scenario with the media data in the first device for collaboration with the AI application of the second device.
- the second device may preset a trigger to enable the AI application coordination function.
- Conditions for example, when the first device establishes a communication connection with the second device, and the second device enables the corresponding AI application, the second device automatically enables the AI application collaboration function.
- the AI application that triggers the AI application collaboration function can be determined by setting a whitelist. As shown in Figure 4b, the whitelist can be set by the user on the interface of the AI application collaboration function, or it can be set by default from the factory. This is not limited.
- the first device may also preset a trigger condition for enabling the AI application coordination function. For example, after the first device establishes a communication connection with the second device, the first device automatically enables the AI application coordination function.
- the user can set the AI application collaboration control interface in the second device: the searched first device has the AI application enabled The first device of the synergy function. Therefore, the second device only establishes a communication connection with the first device with the AI application coordination function enabled, so as to avoid establishing unnecessary communication connections and wasting network resources.
- the second device determines that the first device enables the AI application coordination function, it establishes a communication connection with the first device.
- the first device has the ability to acquire media information and can load the input part of the AI algorithm model; the second device can load the output part of the AI algorithm model; each of the first device and the second device can load one or more AI algorithms The input and output parts of the model.
- the second device may further inquire the model ID and version ID of the loaded AI algorithm model of the first device from the second device through the capability negotiation request message.
- the first device may return a capability negotiation response to the second device, and the capability negotiation response message may include: the model ID of the AI algorithm model loaded by the first device. and version ID.
- the model ID and version ID of the AI algorithm model can be displayed on the AI application collaboration interface, and the loading status is displayed.
- the unit 1 of the first device is used to establish AI application collaboration for the first application with the second device.
- the first device can be displayed in the display bar of the unit 1 Model ID and version ID corresponding to the loaded first application.
- the algorithm model is stored on the second device as an example. It can be determined by judging whether the model ID and version ID corresponding to the input part of the AI algorithm model loaded by the first device are consistent with or compatible with the model ID and version ID corresponding to the output part of the AI algorithm model loaded by the second device.
- the second device determines that the model ID and version ID corresponding to the input part of the AI algorithm model loaded by the first device are consistent with or compatible with the model ID and version ID corresponding to the output part of the AI algorithm model loaded by the second device, It can be confirmed that the establishment of the AI application collaboration of the first device and the second device is completed.
- the second device may determine that the version of the first device needs to be updated, and at this time, the second device may display a loading failure interface on the display screen, and further, may also display a prompt box on the display interface, where the prompt box is used to prompt the user whether to Update the model or version of the model of the first device.
- the second device may display an update interface on the display screen, for example, as shown in (b) of FIG.
- the unit 1 of the first device is used to establish a relationship with the second device.
- the model ID and version ID corresponding to the first application loaded by the first device need to be updated.
- the input part of the AI algorithm model with the model ID corresponding to the output part of the AI algorithm model loaded by the second device and the version ID corresponding to or compatible can be sent to the first device, and the AI algorithm is loaded on the first device.
- the input part of the model can also be displayed in the update interface displayed on the display screen.
- the above description is given by taking the input part and the output part of the AI algorithm model stored in the second device as an example.
- the first device stores the input part and the output part of the AI algorithm model
- the above-mentioned embodiment can also be referred to to complete
- the first device and the second device load the input part and the output part of the AI algorithm model correspondingly.
- the input part and output part of the AI algorithm model can also be stored in the server. At this time, the process of loading the input part and output part of the AI algorithm model correspondingly between the first device and the second device can be completed through the server, which is not repeated here. Repeat.
- the input part and the output part can be in one-to-one correspondence, or multiple input parts can correspond to one output part, that is, an AI algorithm model includes multiple input parts and one output part.
- An input part corresponds to a first device, and an output part corresponds to a second device.
- multiple cameras can be used as multiple first devices, and the device that obtains machine vision processing results is used as the second device.
- the data information obtained by multi-modal sensors such as cameras and lidars is used for applications such as environmental understanding; at this time, the camera can be used as a first device, lidar as a first device, and each sensor as a first device.
- the first device the device that obtains the processing results of applications such as environment understanding is used as the second device.
- multiple media signals on one or more first devices may be processed through multiple AI algorithm model input parts loaded on multiple different neural network processing units. That is, each first device performs feature extraction on the acquired media information, and sends the feature data after feature extraction to the second device, so that the second device can use the acquired media information corresponding to the multiple sensors of the first device. characteristic data.
- multiple feature data obtained on one or more first devices can be independently packaged and packaged, aggregated in the transmission interface and then transmitted uniformly.
- the second device can Packet, restore to determine the feature data corresponding to each first device, synchronize the feature data of each first device and input it to the output part of the AI algorithm model loaded on the neural network processing unit of the second device for processing to obtain The inference results are used for subsequent AI applications, thereby realizing collaborative AI tasks among multiple devices.
- one or more first devices may establish a communication link with the second device respectively, so that each first device may send corresponding feature data to the second device respectively.
- the second device can input the received feature data sent by each first device to the output part of the AI algorithm model corresponding to the corresponding second device for processing, so as to obtain the inference result of the AI algorithm model, and realize multi-device communication. collaborative AI tasks.
- the first device may be a screen end of a split TV, an AR/VR head-mounted display device, or other devices.
- the second device may be a split TV host box, a mobile phone, a PC, a game host, and the like.
- FIG. 5a it is a schematic diagram of the system architecture corresponding to this example.
- the first device and the second device establish a communication connection for AI application collaboration through corresponding transmission interfaces, so as to form a distributed system for AI application collaboration.
- the flow of the media information transmission method may include the following steps:
- Step 501 The first device acquires media information.
- the first device may be hardware capable of capturing, displaying and playing audio and video and processing AI algorithm models.
- the original audio and video signals are collected by the audio and video acquisition unit of the first device, and after preprocessing by the processing unit of the first device, the media information to be transmitted is obtained and transmitted to the NPU of the first device.
- the first device is a screen end of a split TV, and multiple video image frames of a person collected on the camera of the first device are used as media information to be transmitted.
- Step 502 The first device performs feature extraction on the media information according to the input part of the AI algorithm model (for example, the first feature extraction model), and determines feature data.
- the AI algorithm model for example, the first feature extraction model
- the first device determines that the NPU of the first device is loaded with the input part of the AI algorithm model of the first application according to the AI application (for example, application 1) currently determined to be co-processed, so that through the first device in the NPU of the first device, the input part of the AI algorithm model of the first application is loaded.
- the input part of the applied AI algorithm model performs feature extraction on media information to obtain feature data.
- the AI algorithm model corresponding to the first application may be an AI algorithm model for recognizing human gestures.
- the input part of the AI model for gesture recognition performs feature extraction on multiple video image frames of the person collected on the camera of the first device, thereby determining feature data after feature extraction.
- Step 503 The first device sends feature data to the second device.
- the feature data is transferred to the NPU of the second device.
- the second device has media information processing and control capabilities, for example, has AI algorithm model processing hardware and media information processing and human-computer interaction capabilities.
- Step 504 The second device processes the feature data according to the output part of the AI algorithm model (for example, the first feature data processing model).
- the second device can obtain the inference result of the AI algorithm model.
- the inference results of the AI algorithm model are provided to subsequent AI applications to obtain the processing results of subsequent AI application tasks such as voice interaction, machine vision interaction, and environment modeling.
- the processing result of the AI application obtained by the second device needs to be displayed on the display interface of the first device.
- the acquired feature data can be processed according to the output part of the AI model for gesture recognition, so as to recognize the gesture in the video image of the person collected by the first device.
- the recognized gesture can be processed by the AI application by generating a recognition frame at the corresponding position of the image.
- the processing result of the AI application can be displayed on the display screen of the first device.
- the second device can send a control instruction of the gesture recognition result to the first device to instruct the first device on the corresponding display screen of the first device.
- the gesture recognition result is displayed on the position, so that the user can determine that the gesture recognition is successful according to the displayed gesture recognition result.
- step 505a the second device sends the first message to the first device.
- the first message is used to indicate the state of media data collection by the first device, and the state of media data collection by the first device includes at least one of the following: an on state, an off state, or a parameter for collecting media data.
- Step 506a In response to the first message, adjust the state of media data collection by the first device.
- the second device when the second device has the control authority of the camera of the first device, the second device can also determine whether the parameter settings of the camera of the first device are reasonable according to the gesture recognition result.
- a corresponding control command can be generated, and the control command is used to adjust the parameters of the camera of the first device (for example, the pose of the camera of the first device, the focal length of the camera, the focus position, etc., open (type of camera, etc.), in the process of enabling the user to use the machine vision application, it is ensured that the camera of the first device can adjust the parameters of the camera correspondingly with the change of the user's gesture, posture or position, so as to realize the real-time tracking and recognition of the gesture.
- step 505b the second device sends a second message to the first device.
- the second message is used to instruct the first device to obtain the first data.
- Step 506b The first device acquires the first data.
- the first device may acquire the first data based on a network or a storage unit of the first device, or a media information collection unit of the first device may collect the first data. Therefore, the first device may send the first data to the second device; the first data may be: media data collected by the first device, data stored by the first device, and the first device received data.
- the second device when the second device needs to control the media information collection unit of the first device to adjust the parameters of the media information collection unit, it can first send a second message to the first device, and the second message can be used to request to obtain the first device. Parameters of the media information collection unit of the device.
- the second device may control the first device to collect media information based on the reasoning result of the AI application, and in this case, the second message may be used to instruct the first device to collect feature data of the third media data.
- the second device determines that audio information needs to be collected by the first device.
- the second message may be used to instruct the first device to collect the corresponding audio information.
- the second message may also Indicates the model ID and version ID for feature extraction of the audio information after the audio information is collected.
- the first device collects the third media data in response to the second message, and performs feature extraction on the third media data according to the feature extraction model corresponding to the model ID and the version ID to obtain third feature data , and send the third feature data to the second device. Therefore, after receiving the third feature data, the second device processes the third feature data according to the feature data processing model corresponding to the model ID and the version ID, so as to obtain an inference result of the AI application corresponding to the audio information.
- the task of AI application can be better realized.
- the second device may further generate a third message to be sent to the first device according to the received operation information of the user and/or the processing result of the AI model.
- the user may operate the interface of the AI application (eg, application 1), and the operation may be a click operation, a gesture operation, or a voice command operation, which is not limited herein.
- the second device may receive the user's operation instruction on the AI application interface collected from the first device, so as to generate a third message to the first device according to the operation instruction and the processing result of the AI model.
- step 505c the second device sends a third message to the first device.
- the third message is determined by the second device according to the first feature data; the third message is used to indicate the content displayed by the first device.
- Step 506c In response to the third message, the first device displays, through the display unit, the content in the third message that is used to instruct the first device to display.
- the second device recognizes the gesture and determines that the gesture operation instruction is to open the corresponding AI application. At this time, the second device starts the AI according to the operation instruction. application, and send to the first device the media information required to display the startup interface of the AI application. At this time, after receiving the media information, the first device may display an opening interface of the AI application through the display screen of the first device.
- the user's operation instruction is used to jump to the video interface of the corresponding operation and display the corresponding video.
- the second device obtains the media information of the corresponding video according to the recognized user's operation instruction, so as to transmit the corresponding video. interface, sending the media information to the first device, and displaying and playing on the display screen of the first device, so as to realize the response to the user's operation instruction and complete the human-computer interaction.
- the second device has a display screen.
- the first device may be an external camera accessory, a car camera, a home surveillance camera, a smart home appliance with video capture capability, a screen end of a split TV, an AR/VR head-mounted display, and other terminal devices.
- the second device may be a terminal device with strong computing display, such as a split TV host box, a mobile phone, a car host, a PC, and a game host.
- the first device and the second device establish a communication connection for AI application collaboration through corresponding transmission interfaces, so as to form a distributed system for AI application collaboration.
- the flow of the media information transmission method may include the following steps:
- Step 601 The first device acquires media information.
- the first device may be hardware capable of capturing audio and video and processing AI algorithm models.
- the original audio and video signals are collected by the audio and video acquisition unit of the first device, and after preprocessing by the processing unit of the first device, the media information to be transmitted is obtained and transmitted to the NPU of the first device.
- the media information to be transmitted is obtained and transmitted to the NPU of the first device.
- the first device taking the first device as a sensor unit in the vehicle-mounted device as an example, the video images of the road and the outside of the vehicle during the driving of the vehicle collected on the camera of the first device are used as the media information to be transmitted.
- Step 602 The first device performs feature extraction on the media information according to the input part of the AI algorithm model (for example, the first feature extraction model), and determines feature data.
- the AI algorithm model for example, the first feature extraction model
- the NPU of the first device is loaded with the input part of the AI algorithm model, so that feature extraction is performed on the media information through the input part of the AI algorithm model in the NPU to obtain feature data.
- an AI application is an application related to automatic driving.
- the AI algorithm model corresponding to the AI application can be an AI algorithm model used to identify the environment such as the lane where the vehicle is driving, road conditions, etc.
- the input part of the model performs feature extraction on the image of the lane collected on the sensor (eg, radar sensor, camera, etc.) of the first device, so as to determine the feature data after feature extraction.
- the media information collection unit used for AI application collaboration on the first device may be the sensor unit 1 (for example, a radar sensor).
- feature extraction can be performed by using the media information collected by the sensor unit 1 and the corresponding first feature extraction model of the sensor unit 1 to obtain feature data 1 .
- Step 603 The first device sends feature data to the second device.
- the feature data is transferred to the NPU of the second device.
- the second device has media information processing and control capabilities, for example, has AI algorithm model processing hardware, media information processing and human-computer interaction capabilities, and display capabilities.
- Step 604 The second device processes the feature data according to the output part of the AI algorithm model (for example, the first feature data processing model).
- the second device can obtain the inference result of the AI algorithm model.
- the inference results of the AI algorithm model are provided to subsequent AI applications to obtain the processing results of subsequent AI application tasks such as voice interaction, machine vision interaction, and environment modeling.
- step 604 the feature data 1 after feature extraction is performed according to the point cloud image of the lane collected by the radar sensor of the first device is input to the output part of the corresponding AI algorithm model for lane recognition for processing, and the result can be obtained Lane information in the image (as shown in Figure 6c, it can be determined that the vehicle is about to enter the second lane from left to right), so that the second device can be provided with more accurate positioning information of the first device (that is, where the vehicle is located). Lane), so that applications such as a better navigation path can be provided for the first device according to the positioning information of the first device.
- the first device may be various types of media information collection devices, which are used to provide more media information to the AI application, so as to obtain better results of the AI algorithm. Still taking the various types of sensors in the vehicle as an example, for example, under different weather conditions, there may be large errors in identifying the current road only through a single sensor. The received media information can be comprehensively identified, so that a better road identification effect can be obtained.
- the collected media information may be a video/image signal or a combination of multiple video/image signals; each video/image signal may be a visible light image, an infrared image, a radar signal, depth information, etc.
- Modal video/image signal at this time, multiple media information acquisition units on one or more first devices can process the input parts of multiple AI algorithm models loaded by multiple different NPUs, and extract corresponding characteristic data.
- the media information collection unit for AI application collaboration on the first device may include a sensor unit 1 (eg, a radar sensor) and a sensor unit 2 (eg, a camera).
- the feature data 1 can be obtained by performing feature extraction through the media information collected by the sensor unit 1 and the corresponding first feature extraction model of the sensor unit 1 .
- the media information obtained by the sensor is collected by the sensor unit 2 , and feature extraction is performed by using the corresponding first feature extraction model of the sensor unit 2 to obtain feature data 2 . Therefore, multiple feature data output by each NPU can also be independently packaged and encapsulated, aggregated in the transmission interface, and then uniformly transmitted to the second device.
- the second device inputs each feature data 1 and feature data 2 to the output part of the respective AI algorithm model for processing, or inputs to the feature data 1 and feature data 2 for processing. Fusion processed AI algorithm model for better recognition effect.
- Step 605 Display the processing result of the AI application on the second device.
- the processing result of the AI application can be displayed on the second device.
- the lane where the vehicle is currently located may be displayed on the display screen of the second device, and the navigation path planned for the user based on the lane where the vehicle is located may be displayed, and the like.
- the media information collection unit used for AI application collaboration on the first device may include sensor unit 1 (marked as 1 in the figure) and sensor unit 2 (marked as 2 in the figure).
- a control instruction for the first device may also be generated.
- the first device and the second device are connected through a transmission interface to form a distributed system for AI application collaboration, which combines the information perception capability and lightweight AI processing capability of the first device with the more powerful computing hardware, AI processing capability and interaction of the second device.
- the combination of capabilities can collaboratively complete tasks of AI applications such as voice interaction, visual interaction, and environment modeling.
- the first device and the second device can form a distributed voice interaction system for AI application collaboration.
- the first device may have audio information collection capability and AI algorithm model processing hardware.
- the second device may be a terminal device with strong computing, such as a mobile phone, a smart TV, and a vehicle-mounted host.
- the second device may also have a display function.
- the AI application cooperates with the corresponding transmission interface, which can be a wireless transmission system such as wifi and Bluetooth, or an electrical signal transmitted by wire or an optical signal transmitted by optical fiber.
- the present application provides a schematic flowchart of a method for transmitting media information, which specifically includes:
- Step 701 The first device acquires media information.
- the first device may be hardware capable of collecting audio and processing an AI algorithm model.
- the original audio signal is collected by the audio unit of the first device, and after preprocessing by the processing unit of the first device, the media information to be transmitted is obtained and transmitted to the NPU of the first device.
- the first device is a smart earphone, and AI application collaboration (for example, AI applications such as noise reduction, voice interaction, etc.) is established with the second device through the microphone unit in the first device.
- Media information such as voice input by the user or environmental noise collected on the Internet.
- Step 702 The first device performs feature extraction on the media information according to the input part of the AI algorithm model (for example, the first feature extraction model), and determines feature data.
- the AI algorithm model for example, the first feature extraction model
- the NPU of the first device is loaded with the input part of the AI algorithm model, so that feature extraction is performed on the media information through the input part of the AI algorithm model in the NPU to obtain feature data.
- the AI application is an application related to speech recognition interaction.
- the AI algorithm model corresponding to the AI application may be the AI algorithm model used for speech recognition.
- Feature extraction is performed on the audio information collected on the microphone of the first device, thereby determining feature data after feature extraction.
- the AI application is an application related to automatic noise reduction.
- the AI algorithm model corresponding to the AI application may be the AI algorithm model used to identify environmental noise.
- the input part of the AI model for noise identification can be used. , perform feature extraction on the audio information collected on the microphone of the first device, thereby determining the feature data after feature extraction.
- Step 703 The first device sends feature data to the second device.
- the feature data is transferred to the NPU of the second device.
- the second device has media information processing and control capabilities, for example, has AI algorithm model processing hardware, media information processing and human-computer interaction capabilities, and display capabilities.
- Step 704 The second device processes the feature data according to the output part of the AI algorithm model (for example, the first feature data processing model) to determine the result of the first application.
- the AI algorithm model for example, the first feature data processing model
- step 705 the second device sends the result of the first application to the first device.
- Step 706 The first device plays the result of the first application.
- the second device can obtain the inference result of the AI algorithm model.
- the inference results of the AI algorithm model are provided to subsequent AI applications to obtain the processing results of subsequent voice interaction tasks such as speech recognition, natural language processing, and voiceprint recognition.
- the voice recognition result collected by the first device may be determined according to the processing of the feature data, for example, the voice recognition result is a video specified by the search.
- the second device can display the speech recognition result on the display interface of the second device, and search for the corresponding video according to the speech recognition result. Go to the corresponding video playback application to display the searched video to complete the task of the AI application of voice recognition interaction.
- the noise recognition result collected by the first device can be determined according to the processing of the feature data, so as to generate corresponding noise reduction audio information according to the noise recognition result, and send the noise reduction audio information to the first device through the transmission interface.
- a device so that when the microphone of the first device performs recording or the audio and video playback unit performs audio playback, the noise reduction of the audio recording or audio playback of the first device is realized through the noise reduction audio information.
- the audio information obtained by the first device can not be transmitted to the second device, but is converted into abstract feature data after the input part of the AI algorithm model and then transmitted; the feature data has obvious information loss after model processing, which cannot be recovered by human beings.
- the audio and video information that can be directly understood improves the privacy protection ability; the data volume of the feature data is significantly lower than the original audio and video information, and it can also be transmitted in real time under the condition of small channel bandwidth, which saves the extra compression and coding process and reduces the system cost. Power consumption and delay, reduce costs and improve product competitiveness.
- the interface does not require It supports bidirectional transmission of video information, requires low bandwidth, and is more suitable for short-distance wireless transmission.
- the methods provided by the embodiments of the present application are introduced from the perspective of the first device and the second device as execution subjects.
- the electronic device may include a hardware structure and/or software modules, and implement the above functions in the form of a hardware structure, a software module, or a hardware structure plus a software module. Whether one of the above functions is performed in the form of a hardware structure, a software module, or a hardware structure plus a software module depends on the specific application and design constraints of the technical solution.
- FIG. 8 shows an electronic device 800 of the present application, including a transceiver module 801, a collection module 803 and a processing module 802.
- the electronic device 800 may further include a display module.
- the electronic device 800 may be the first device in this embodiment of the present application.
- the transceiver module 801 includes a first transmission interface.
- a collection module 803, configured to collect first media information
- the processing module 802 is configured to perform feature extraction on the first media information, and determine the first feature data of the first media information; send the first feature data to the second device through the first transmission interface, the The first feature data is used for the second device to obtain a result of the first application.
- the transceiver module 801 is configured to receive a capability negotiation request message sent by the second device through the first transmission interface; the capability negotiation request message is used to request a transmission supported by the first device protocol, and the feature extraction capability of the first device; the transmission protocol of the first device is used to instruct the first device to support the transmission of feature data; the feature extraction capability of the first device is used to instruct the first device The device supports extracting the first feature data of the first media information; sends a capability negotiation response message to the second device through the first transmission interface; the capability negotiation response message is used to confirm that the first device supports transmission The transmission protocol of the feature data and the feature extraction capability of the first device.
- the processing module 802 is configured to, in response to the first operation on the first application, send a first notification message to the second device through the transceiver module 801;
- the second device establishes an electronic device for communication connection;
- the first notification message is used to request the first device to establish a first application collaboration with the second device;
- the transceiver module 801 is configured to receive the returned message from the second device A first response message;
- the first response message is used to confirm that the first device and the second device start the first application collaboration.
- the transceiver module 801 is configured to receive a first notification message sent from the second device; the first device is an electronic device that establishes a communication connection with the second device; the first notification The message is used to request the first device to establish a first application collaboration with the second device; the processing module 802 is configured to send the message to the second device through the transceiver module 801 in response to a third operation on the first application A first response message; the first response message is used to confirm that the first device and the second device start the first application collaboration.
- the transceiver module 801 is configured to send a capability negotiation request message to the second device through the first transmission interface; the capability negotiation request message is used to request a transmission supported by the second device protocol, and the feature data processing capability of the second device, the transmission protocol of the second device is used to instruct the second device to support the transmission of feature data; the feature data processing capability of the second device is used to instruct the second device
- the second device supports the ability to process the first feature data to obtain the result of the first application; receives a capability negotiation response message from the second device through the first transmission interface; the capability negotiation response message is used to confirm the
- the second device supports a transmission protocol for transmitting feature data and the feature data processing capability of the second device.
- the processing module 802 is configured to obtain a first feature extraction model through the transceiver module 801; wherein, the first feature extraction model is used to perform feature extraction on the first media information, and the first feature extraction model
- the version of the feature extraction model corresponds to the version of the first feature data processing model, and the first feature data processing model is used by the second device to process the first feature data to obtain the result of the first application.
- the capability negotiation response message further includes: the version of the feature extraction model in the first device; or the version of the feature data processing model in the second device.
- the transceiver module 801 is configured to receive the first feature extraction model from the second device through the first transmission interface, or receive the first feature extraction model from the server, or read The first feature extraction model stored by the first device.
- the transceiver module 801 is configured to send the first feature data processing model to the second device through the first transmission interface; wherein the version of the first feature extraction model is the same as the first feature extraction model.
- the version of the feature data processing model corresponds, and the first feature data processing model is used by the second device to process the first feature data to obtain the result of the first application.
- the processing module 802 is configured to obtain a second feature extraction model through the transceiver module 801, the version of the second feature extraction model corresponds to the version of the second feature data processing model, and the second feature extraction model corresponds to the version of the second feature data processing model.
- the model and the second feature data processing model are determined after updating the first feature extraction model and the second feature data processing model.
- the processing module 802 is configured to perform feature extraction on the training samples according to the first feature extraction model to generate first training feature data;
- the second device sends the first training feature data;
- the first training feature data is used to train the first feature extraction model and the first feature data processing model.
- the transceiver module 801 is configured to receive feedback data from the second device through the first transmission interface, where the feedback data is trained by the second device according to the first training feature data determined later; the feedback data is used by the first device to train the first feature extraction model.
- the transceiver module 801 is configured to receive a first message from the second device through the first transmission interface; the first message is used to indicate the state of media information collection by the first device ; In response to the first message, adjust the state of the media information collected by the first device.
- the state in which the first device collects media information includes at least one of the following: an on state, an off state, or a parameter for collecting media information.
- the transceiver module 801 is configured to receive a second message from the second device through the first transmission interface; wherein the second message is used to instruct the first device to obtain the first data; the processing module 802 is configured to, in response to the second message, obtain the first data, or collect the first data; send the first data to the second device; the first data is one of the following: media information collected by the first device, parameters of the first device, data stored by the first device, and data received by the first device.
- the transceiver module 801 is configured to send the first data to the second device through the first transmission interface.
- the transceiver module 801 is configured to receive a second message from the second device through the first transmission interface, where the second message is used to instruct the first device to collect third media information feature data of the device; collect the third media information in response to the second message; perform feature extraction on the third media information to obtain third feature data; send the third media information to the second device through the first transmission interface The third characteristic data is sent.
- the second message or the first message is determined by the second device according to the first feature data.
- the transceiver module 801 is configured to receive a third message from the second device through the first transmission interface, where the third message is the second device according to the first feature data It is determined that the third message is used to indicate the content displayed by the first device; the processing module 802 is configured to, in response to the third message, display the third message through the display module to indicate the first Content displayed by a device.
- the transceiver module 801 is configured to receive, through the first transmission interface, an authentication request message sent by the second device, where the authentication request message is used to request whether the first device communicates with the first device.
- the two devices establish a communication connection, and the communication connection is used to confirm the authority of the second device to control the first device; send an authentication response message to the second device through the first transmission interface; the authentication response message It is used to confirm the authority of the second device to control the first device.
- the transceiver module 801 is configured to receive an authentication success message sent by the second device through the first transmission interface; the authentication success message includes: a device identifier corresponding to the first device, and The identifier of the distributed system where the first device and the second device are located.
- the electronic device 800 may further include a first module; the authentication success message further includes at least one of the following: an identifier of the first module of the first device, and the first module in the Identity in a distributed system.
- the transceiver module 801 may further include a third transmission interface; the first device and the second device establish a channel connection through the third transmission interface; the characteristic data or message sent by the first device is: After being encapsulated into the first bit stream data through the first transmission interface, the data is sent through the third transmission interface.
- the first device and the second device establish a channel connection through a third transmission interface; the message received by the first device is the second bit stream data received through the third transmission interface , and obtained by decapsulating the second bit stream data through the first transmission interface.
- FIG. 9 shows an electronic device 900 according to the present application, including a transceiver module 901 and a processing module 902.
- the electronic device 900 may further include a display module.
- the electronic device 900 may be the second device in this embodiment of the present application.
- the transceiver module 901 includes a second transmission interface.
- a transceiver module 901 configured to receive first feature data from a first device through the second transmission interface; the first feature data is determined according to feature extraction of the collected first media information by the first device;
- the processing module 902 is configured to process the first feature data to obtain a processing result of the first application.
- the processing module 902 is configured to, in response to the second operation on the first application, send a first notification message to the first device through the transceiver module 901;
- the first notification message is used to request the first device to establish a first application collaboration with the second device;
- the first response message is used to confirm that the first device and the second device start the first application collaboration.
- the processing module 902 is configured to receive, through the transceiver module 901, a first notification message sent by a first device; the first device is an electronic device that establishes a communication connection with the second device; the first device A notification message is used to request the first device to establish a first application collaboration with the second device; in response to a fourth operation on the first application, send a first response message to the first device through the transceiver module 901 ; the first response message is used to confirm that the first device and the second device enable the first application collaboration.
- a capability negotiation request message is sent to the first device through the second transmission interface; the capability negotiation request message is used to request a transmission protocol supported by the first device, and the first device
- the feature extraction capability of the device is used to instruct the first device to support the transmission of feature data; the feature extraction capability of the first device is used to instruct the first device to support the extraction of the first device
- the first feature data of the media information is received through the second transmission interface; the capability negotiation response message is used to confirm that the first device supports a transmission protocol for transmitting feature data.
- a transceiver module 901 configured to receive a capability negotiation request message sent by the first device through the second transmission interface; the capability negotiation request message is used to request a transmission protocol supported by the second transmission interface, and the second transmission interface
- the feature data processing capability of the second device the transmission protocol of the second device is used to instruct the second device to support the transmission of feature data; the feature data processing capability of the second device is used to instruct the second device to support processing all the first feature data to obtain the capability of the result of the first application; send a capability negotiation response message to the first device through the second transmission interface; the capability negotiation response message is used to confirm that the second device supports the transmission feature The data transmission protocol and the characteristic data processing capability of the second device.
- the processing module 902 is used to obtain a first feature data processing model; the first feature data processing model is used by the second device to process the first feature data to obtain the first feature data.
- the capability negotiation response message further includes: the version of the feature extraction model in the first device; or the version of the feature data processing model in the second device.
- the transceiver module 901 is configured to receive the first feature data processing model from the first device through the second transmission interface, or receive the first feature data processing model from the server, or, A processing module 902, configured to read the first feature data processing model stored in the second device;
- the transceiver module 901 is configured to send the first feature extraction model to the first device through the second transmission interface; wherein, the version of the first feature extraction model is processed with the first feature data The version of the model corresponds, and the first feature data processing model is used by the second device to process the first feature data to obtain the result of the first application.
- the processing module 902 is configured to obtain a second feature data processing model, the version of the second feature data processing model corresponds to the version of the second feature extraction model, and the second feature extraction model and the second feature data processing model is determined after updating the first feature extraction model and the second feature data processing model.
- the processing module 902 is configured to receive the first training feature data through the transceiver module 901; the first training feature data is that the first device features the training samples according to the first feature extraction model. Determined after extraction; training the first feature data processing model according to the first training feature data.
- the processing module 902 is used to obtain feedback data of the first feature extraction model; the feedback data is determined by the second device after training according to the first training feature data; the feedback data It is used for the first device to train the first feature extraction model; the feedback data is sent to the first device through the transceiver module 901 .
- a possible implementation manner is to receive the second feature data sent by the second device through the transceiver module 901; the second feature data is extracted by the first device according to the second media information collected and the second feature The model is determined after feature extraction.
- the processing module 902 is configured to process the second feature data according to the second feature data processing model to obtain the result of the first application.
- the transceiver module 901 is configured to send a first message to the first device through the second transmission interface; the first message is used to indicate the state of media information collection by the first device.
- the state in which the first device collects media information includes at least one of the following: an on state, an off state, or a parameter for collecting media information.
- the transceiver module 901 is configured to send a second message to the first device through the second transmission interface; the second message is used to instruct the first device to obtain the first data; the The first data is one of the following: media information collected by the first device, parameters of the first device, data stored by the first device, and data received by the first device.
- the transceiver module 901 is configured to receive the first data from the first device through the second transmission interface.
- the transceiver module 901 is configured to send a second message to the first device through the second transmission interface; the second message is used to instruct the first device to collect feature data of the third media information; The third feature data sent from the first device is received through the second transmission interface; the third feature data is determined after the first device performs feature extraction on the collected third media information.
- the first message or the second message is determined according to a processing result of the first feature data.
- the processing module 902 is configured to generate a third message in response to the processing result of the first feature data; the third message is used to indicate the content displayed by the first device.
- the number of the first devices is N; the method further includes:
- the transceiver module 901 is configured to receive a fourth message through the second transmission interface; the fourth message includes M first characteristic data of the N first devices; N and M are positive integers greater than 1; M is greater than or equal to N;
- the processing module 902 is configured to process the M pieces of first characteristic data according to the characteristic data processing model corresponding to the M pieces of first characteristic data to obtain the result of the first application.
- the transceiver module 901 is configured to send an authentication request message to the first device through the second transmission interface, where the authentication request message is used to request whether the first device communicates with the second device.
- the device establishes a communication connection; the communication connection is used to confirm the authority of the second device to control the first device; the authentication response message sent by the second device is received through the second transmission interface; the authentication response message It is used to confirm whether the first device establishes a communication connection with the second device.
- the processing module 902 is configured to, in response to the authentication response message sent by the second device, set the device identifier corresponding to the first device for the first device, and the first device and the The identifier of the distributed system where the second device is located; the device identifier corresponding to the first device and the identifier of the distributed system are used for the first device to communicate with the second device; the first device The device sends an authentication success message to the second device through the first transmission interface; the authentication success message includes: the device identifier corresponding to the first device, and the location where the first device and the second device are located. The identity of the distributed system.
- the second device includes a second module; the authentication success message further includes at least one of the following: an identifier of the second module, and the second module in the distributed system 's identification.
- the transceiver module 901 further includes a third transmission interface; the first device and the second device establish a channel connection through the third transmission interface; the message sent by the second device is through the third transmission interface. After the second transmission interface is encapsulated into the second bit stream data, the data is sent through the third transmission interface.
- the first device and the second device establish a channel connection through a third transmission interface; the feature data or message received by the second device is the first device received through the third transmission interface. bit stream data, and obtained by decapsulating the second bit stream data through the second transmission interface.
- Embodiments of the present application further provide a media information transmission system, including the electronic device 800 as shown in FIG. 8 or the first device as shown in FIG. 3a , and the electronic device 900 as shown in FIG. 9 or the electronic device 900 as shown in FIG. 3b the second device shown.
- Embodiments of the present application further provide a computer storage medium, where the computer-readable storage medium is used to store a computer program, and when the computer program runs on a computer, the computer can execute any The method described in a possible embodiment.
- the embodiments of the present application further provide a computer program product including instructions, the computer program product is used to store a computer program, and when the computer program is run on a computer, the computer is made to execute any one of the steps in Figs. The method described in a possible embodiment.
- processors mentioned in the embodiments of the present application may be a CPU, and may also be other general-purpose processors, digital signal processors (digital signal processors, DSPs), application specific integrated circuits (application specific integrated circuits, ASICs), ready-made Field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
- DSPs digital signal processors
- ASICs application specific integrated circuits
- FPGA Field programmable gate array
- a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
- the memory mentioned in the embodiments of the present application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.
- the non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically programmable Erase programmable read-only memory (electrically EPROM, EEPROM) or flash memory.
- Volatile memory may be random access memory (RAM), which acts as an external cache.
- RAM random access memory
- SRAM static random access memory
- DRAM dynamic random access memory
- SDRAM synchronous DRAM
- SDRAM double data rate synchronous dynamic random access memory
- double data rate SDRAM double data rate SDRAM
- DDR SDRAM enhanced synchronous dynamic random access memory
- ESDRAM enhanced synchronous dynamic random access memory
- SCRAM synchronous link dynamic random access memory
- direct rambus RAM direct rambus RAM
- the processor is a general-purpose processor, DSP, ASIC, FPGA or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components
- the memory storage module
- memory described herein is intended to include, but not be limited to, these and any other suitable types of memory.
- the size of the sequence numbers of the above-mentioned processes does not mean the sequence of execution, and the execution sequence of each process should be determined by its functions and internal logic, and should not be dealt with in the embodiments of the present application. implementation constitutes any limitation.
- the disclosed system, apparatus and method may be implemented in other manners.
- the apparatus embodiments described above are only illustrative.
- the division of the units is only a logical function division. In actual implementation, there may be other division methods.
- multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented.
- the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.
- the units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.
- each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
- the functions, if implemented in the form of software functional units and sold or used as independent products, may be stored in a computer-readable storage medium.
- the technical solution of the present application can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution.
- the computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.
- the aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program codes .
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Security & Cryptography (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Signal Processing (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Computer And Data Communications (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
本申请提供一种媒体信息传输方法及装置,应用于第一设备时,第一设备包括第一传输接口,该方法包括采集第一媒体信息;对所述第一媒体信息进行特征提取,确定所述第一媒体信息的第一特征数据;通过所述第一传输接口向第二设备发送所述第一特征数据,所述第一特征数据用于所述第二设备获得第一应用的结果。从而,降低传输媒体信息时的传输开销及编解码的开销,提高传输效果。
Description
相关申请的交叉引用
本申请要求在2020年08月14日提交中国专利局、申请号为202010819860.3、申请名称为“面向分布式AI的传输接口系统”的中国专利申请的优先权,其全部内容通过引用结合在本申请中;本申请要求在2020年11月19日提交中国专利局、申请号为202011300677.9、申请名称为“一种媒体信息传输方法及装置”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
本申请实施例涉及媒体技术领域,尤其涉及一种媒体信息传输方法及电子设备。
在机器视觉、语音交互、人脸识别、自动驾驶、环境建模等机器算法应用场景中,图像采集设备采集原始的视频或图像资源,通过将视频或图像资源压缩成视频流,传输给相应的服务器或实现机器算法应用的设备,这些设备对接收到的视频流进行解码或解压后恢复出标准的音视频信号;之后利用深度学习等机器算法对音视频信号进行处理,得到人工智能(artificial intelligence,AI)应用的处理结果。
上述过程中,需要占用大量的带宽,并对机器算法应用的设备而言,也需要大量的解码或解压缩的过程,导致计算资源的浪费和传输带宽资源的浪费。
发明内容
本申请提供了一种媒体信息传输方法及电子设备,用以解决现有技术中采用媒体数据传输过程中,占用较大的带宽资源,不利于媒体资源的后续应用等问题。
第一方面,本申请提供一种媒体信息传输方法,应用于第一设备,所述第一设备包括第一传输接口,所述方法包括:采集第一媒体信息;对所述第一媒体信息进行特征提取,确定所述第一媒体信息的第一特征数据;通过所述第一传输接口向第二设备发送所述第一特征数据,所述第一特征数据用于所述第二设备获得第一应用的结果。
其中,第一应用可以为AI应用,执行上述方法的设备可以是第一设备,也可以是具有相应功能的模块或部件,例如,芯片。下面以第一设备执行该方法为例,第一设备可以为具有采集媒体信息能力和特征提取能力的设备。第一设备可以包括第一传输接口,该第一传输接口支持传输特征数据,从而,实现将媒体信息通过特征数据的方式传输给第二设备,避免了现有技术中需要在发送端对媒体信息进行压缩编码,并需要在媒体信息的接收端进行恢复传输的媒体信息,编解码过程较为复杂,降低了计算资源的开销,并整体降低系统的延迟,有助于应用于实时性要求高的AI应用中,提升用户使用AI应用的体验。另外,由于传输的是抽象后的特征数据,在降低编码难度的同时,也极大的减少了传输的信息量,降低了传输资源的开销。进一步的,考虑到传输媒体信息时的安全性,传输抽象后的特征数据,相比现有技术中传输媒体信息的方式,传输接口内传输的特征数据无法逆向 转换为原始的媒体信息,实现更好的隐私保护能力。
一种可能的实现方式,所述向第二设备发送所述第一特征数据之前,还包括:
通过所述第一传输接口接收所述第二设备发送的能力协商请求消息;所述能力协商请求消息用于请求所述第一设备支持的传输协议,及所述第一设备的特征提取能力;所述第一设备的传输协议用于指示所述第一设备支持传输特征数据;所述第一设备的特征提取能力用于指示所述第一设备支持提取所述第一媒体信息的第一特征数据;通过所述第一传输接口向所述第二设备发送能力协商响应消息;所述能力协商响应消息用于确认所述第一设备支持传输特征数据的传输协议及所述第一设备的特征提取能力。
一种可能的实现方式,响应于在第一应用上的第一操作,向所述第二设备发送第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;接收所述第二设备返回的第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启第一应用协同。
通过上述方法,可以通过第一设备的第一应用上的第一操作,触发第一设备和第二设备建立第一应用协同的过程,并通过第二设备确认是否开启第一应用协同。第一应用协同可以是第一设备通过所述第一传输接口向第二设备发送所述第一特征数据,第二设备根据第一特征数据获得第一应用的结果的协同过程,提升用户使用第一设备和第二设备协同处理第一应用的体验。
一种可能的实现方式,接收来自所述第二设备发送第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;响应于在第一应用上的第三操作,向所述第二设备发送第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启第一应用协同。
通过上述方法,通过第二设备触发第一设备和第二设备建立第一应用协同的建立过程,并根据通过第一设备上第一应用的第三操作,确认是否开启第一应用协同,提升用户使用第一设备和第二设备协同处理第一应用的体验。
一种可能的实现方式,所述向第二设备发送所述第一特征数据之前,还包括:
通过所述第一传输接口向所述第二设备发送的能力协商请求消息;所述能力协商请求消息用于请求所述第二设备支持的传输协议,及所述第二设备的特征数据处理能力,所述第二设备的传输协议用于指示所述第二设备支持传输特征数据;所述第二设备的特征数据处理能力用于指示所述第二设备支持处理所述第一特征数据获得第一应用的结果的能力;通过所述第一传输接口接收来自所述第二设备的能力协商响应消息;所述能力协商响应消息用于确认所述第二设备支持传输特征数据的传输协议及所述第二设备的特征数据处理能力。
通过上述方法,可以通过能力协商的方式,例如,第一设备发起协商请求消息,或者第二设备发起协商请求消息,从而,确认第一设备是否具备特征提取能力和传输特征数据的能力,及第二设备是否支持接收特征数据的能力,及处理特征数据的能力,使得第一设备和第二设备确认是否支持传输特征数据并协同实现第一应用的相应功能,提升AI应用的性能。
一种可能的实现方式,第一设备对所述第一媒体信息进行特征提取之前,还包括:
获取第一特征提取模型;其中,所述第一特征提取模型用于对所述第一媒体信息进行特征提取,所述第一特征提取模型的版本与第一特征数据处理模型的版本对应,所述第一 特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果。
通过上述方法,可以在第一设备和第二设备建立连接,并确定可以执行相应的协同处理的第一应用的任务后,第一设备和第二设备分别加载该第一应用对应的算法模型的输入部分(第一特征数据提取模型)和输出部分(第一特征数据处理模型)。从而实现第一设备和第二设备的第一应用协同。
一种可能的实现方式,所述能力协商响应消息还包括:
所述第一设备中的特征提取模型的版本;或者,所述第二设备中的特征数据处理模型的版本。
通过上述方法,还可以通过能力协商响应,确认第一设备中的特征提取模型的版本和第二设备中的特征数据处理模型的版本是否可以完成第一应用的协同。
一种可能的实现方式,所述获取第一特征提取模型,包括:
通过所述第一传输接口接收来自所述第二设备的第一特征提取模型,或者,接收来自服务器的第一特征提取模型,或者,读取所述第一设备存储的第一特征提取模型。
通过上述方法,可以通过多种方式,获取第一特征提取模型,实现更多灵活的AI应用协同方式。
一种可能的实现方式,所述方法还包括:
通过所述第一传输接口向所述第二设备发送第一特征数据处理模型;
其中,所述第一特征提取模型的版本与所述第一特征数据处理模型的版本对应,所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果。
第一设备可以存储有第一特征数据处理模型,通过上述方法,通过所述第一传输接口向所述第二设备发送第一特征数据处理模型,以使得第一设备和第二设备可以实现AI应用协同处理媒体信息。
一种可能的实现方式,所述方法还包括:
获取第二特征提取模型,所述第二特征提取模型的版本与第二特征数据处理模型的版本对应,所述第二特征提取模型和所述第二特征数据处理模型为更新所述第一特征提取模型和所述第二特征数据处理模型后确定的。
通过上述方法,可以在第一特征提取模型更新后,获取更新后的特征提取模型(即第二特征提取模型),从而,可以适应各种不同AI应用的需求,提高第一设备和第二设备协同处理媒体信息,用于AI应用的适用性。
一种可能的实现方式,所述方法还包括:
根据所述第一特征提取模型,对训练样本进行特征提取,生成第一训练特征数据;
通过所述第一传输接口向所述第二设备发送所述第一训练特征数据;所述第一训练特征数据用于训练所述第一特征提取模型和所述第一特征数据处理模型。
通过上述方法,可以利用第一设备和第二设备进行联合训练,合理利用第一设备和第二设备的算力,提升AI应用的性能。
一种可能的实现方式,通过所述第一传输接口接收来自所述第二设备的反馈数据,所述反馈数据为所述第二设备根据所述第一训练特征数据训练后确定的;所述反馈数据用于所述第一设备训练所述第一特征提取模型。
通过上述方法,通过第二设备反馈的反馈数据,对第一设备上的特征提取模型进行训练,利用第一设备和第二设备进行联合训练的训练效果,提升特征提取模型的性能。
一种可能的实现方式,所述方法还包括:
通过所述第一传输接口接收来自所述第二设备的第一消息;所述第一消息用于指示所述第一设备采集媒体信息的状态;响应于所述第一消息,调整所述第一设备采集媒体信息的状态。
通过上述方法,第一设备可以根据第二设备发送的第一消息,调整第一设备采集媒体信息的状态,以更好的获得第一应用所需采集的媒体信息,提升第一应用的效果。
一种可能的实现方式,所述第一设备采集媒体信息的状态包括以下至少一项:开启状态、关闭状态或采集媒体信息的参数。
一种可能的实现方式,所述方法还包括:通过所述第一传输接口接收来自所述第二设备的第二消息;其中,所述第二消息用于指示所述第一设备获取第一数据;响应于所述第二消息,获取所述第一数据,或者,采集所述第一数据;向所述第二设备发送所述第一数据;所述第一数据为以下一项:所述第一设备采集到的媒体信息,所述第一设备的参数,所述第一设备存储的数据,第一设备接收的数据。
通过上述方法,第二设备可以指示第一设备获取第一数据,使得在第一设备和第二设备之间传输特征数据的同时,兼容其他数据的传输,有助于提升第一设备和第二设备的实现AI应用功能所需的传输性能和传输场景的适应性。
一种可能的实现方式,通过所述第一传输接口向所述第二设备发送所述第一数据。
通过上述方法,第一传输接口可以支持多种数据的传输,提升传输性能和传输场景的适应性。
一种可能的实现方式,通过所述第一传输接口接收来自所述第二设备的第二消息,所述第二消息用于指示所述第一设备采集第三媒体信息的特征数据;响应于所述第二消息,采集所述第三媒体信息;对所述第三媒体信息进行特征提取,得到第三特征数据;通过所述第一传输接口向所述第二设备发送所述第三特征数据。
通过上述方法,可以通过第二设备发送第二消息,以控制第一设备采集媒体信息,并传输相应的第三特征数据,例如,可以通过对第一应用的处理结果或AI应用的需要,确定需采集的第三媒体信息,从而灵活的调整第一设备采集的媒体信息,以使AI应用可以获得更好的结果,从而,整体提高AI应用的效果。
一种可能的实现方式,所述第二消息或所述第一消息为所述第二设备根据所述第一特征数据确定的。
通过上述方法,第二设备可以基于第一设备传输的第一特征数据,确定第一应用的结果,并根据第一应用的结果生成第一消息或第二消息,从而向第一设备反馈相应的第一消息或第二消息,第一设备可以响应于第一消息或第二消息,调整媒体信息的采集、获取和传输,以使第一设备和第二设备更好的完成第一应用协同。
一种可能的实现方式,所述第一设备还可以包括显示单元;所述方法还包括:通过所述第一传输接口接收来自所述第二设备的第三消息,所述第三消息为所述第二设备根据所述第一特征数据确定的,所述第三消息用于指示所述第一设备显示的内容;响应于所述第三消息,通过显示单元显示所述第三消息中用于指示所述第一设备显示的内容。
通过上述方法,可以通过第三消息,获得待显示的内容,该内容可以是第一应用的处 理结果,也可以是其他第二设备需要第一设备显示的内容,从而,使得第一设备和第二设备更好的实现AI应用的协同,提升AI应用的使用体验。
一种可能的实现方式,所述方法还包括:
通过所述第一传输接口接收所述第二设备发送的认证请求消息,所述认证请求消息用于请求所述第一设备是否与所述第二设备建立通信连接,所述通信连接用于确认所述第二设备控制所述第一设备的权限;
通过所述第一传输接口向所述第二设备发送认证响应消息;所述认证响应消息用于确认所述第二设备控制所述第一设备的权限。
通过上述方法,可以通过第一设备和第二设备之间的认证,确认第二设备是否可以获得控制第一设备的权限,从而,在第二设备根据第一特征数据获得第一应用的结果后,调整第一设备采集媒体信息,有助于获取更好的第一应用的结果,提升AI应用的性能。
一种可能的实现方式,所述方法还包括:
通过所述第一传输接口接收所述第二设备发送的认证成功消息;所述认证成功消息包括:所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识。
通过上述方法,还可以将第一设备和第二设备设置为分布式系统中的设备,以实现对第一设备和第二设备更好的管理,有利于利用多个设备实现AI应用协同。
一种可能的实现方式,所述第一设备包括第一模块;所述认证成功消息还包括以下至少一项:所述第一设备的第一模块的标识,及所述第一模块在所述分布式系统中的标识。
通过上述方法,还可以将第一设备中的模块设置为分布式系统中的模块,从而,为第二设备控制各设备中的模块,协同完成AI应用做好准备。
一种可能的实现方式,所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第一设备发送的特征数据或消息为通过所述第一传输接口封装为第一比特流数据后,通过所述第三传输接口发送的。
通过上述方法,通过第一传输接口封装特征数据,并通过第三传输接口向第二设备发送封装后的数据,从而,通过第三传输接口,可以兼容多种传输协议,也可以实现聚合传输等功能,从而提升第一设备的传输能力和传输媒体信息的兼容性。
一种可能的实现方式,所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第一设备接收的消息为通过所述第三传输接口接收的第二比特流数据,并通过所述第一传输接口将所述第二比特流数据解封装后得到的。
通过上述方法,通过第一传输接口解封装第三传输接口接收的来自第二设备的数据,通过第三传输接口,可以兼容多种传输协议,也可以实现聚合传输等功能,从而提升第一设备的传输能力和传输媒体信息的兼容性。
第二方面,本申请提供一种媒体信息传输方法,应用于第二设备;所述第二设备包括第二传输接口;所述方法包括:通过所述第二传输接口接收来自第一设备的第一特征数据;所述第一特征数据为根据第一设备对采集的第一媒体信息进行特征提取后确定的;对所述第一特征数据进行处理,得到第一应用的处理结果。
其中,第一应用可以为AI应用,执行上述方法的设备可以是第二设备,也可以是具有相应功能的模块或部件,例如,芯片。下面以第二设备执行该方法为例,第二设备可以为具有接收特征数据能力和对特征数据处理能力的设备。第二设备可以包括第二传输接口, 该第二传输接口支持传输特征数据,从而,实现第二设备接收特征数据,而不是直接接收媒体数据,避免了现有技术中需要在发送端对媒体信息进行压缩编码,并需要在媒体信息的接收端进行恢复传输的媒体信息,编解码过程较为复杂,降低了计算资源的开销,并整体降低系统的延迟,有助于应用于实时性要求高的AI应用中,提升用户使用AI应用的体验。另外,由于传输的是抽象后的特征数据,在降低编码难度的同时,也极大的减少了传输的信息量,降低了传输资源的开销。进一步的,考虑到传输媒体信息时的安全性,传输抽象后的特征数据,相比现有技术中传输媒体信息的方式,传输接口内传输的特征数据无法逆向转换为原始的媒体信息,实现更好的隐私保护能力。
一种可能的实现方式,响应于在第一应用上的第二操作,向所述第一设备发送第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;接收所述第二设备返回的第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启第一应用协同。
通过上述方法,通过第二设备响应在第一应用上的第二操作,触发第一设备和第二设备建立第一应用协同的建立过程,并根据通过第一设备确认是否开启第一应用协同,提升用户使用第一设备和第二设备协同处理第一应用的体验。
一种可能的实现方式,接收第一设备发送的第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;响应于在第一应用上的第四操作,向所述第一设备发送第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启第一应用协同。
通过上述方法,可以通过第一设备触发第一设备和第二设备建立第一应用协同的建立过程,并通过第二设备响应在第一应用上的第四操作,确认是否开启第一应用协同。第一应用协同可以是第一设备通过所述第一传输接口向第二设备发送所述第一特征数据,第二设备根据第一特征数据获得第一应用的结果的协同过程,提升用户使用第一设备和第二设备协同处理第一应用的体验。
一种可能的实现方式,所述接收来自第一设备的第一特征数据之前,还包括:
通过所述第二传输接口向所述第一设备发送能力协商请求消息;所述能力协商请求消息用于请求所述第一设备支持的传输协议,及所述第一设备的特征提取能力;所述第一设备的传输协议用于指示所述第一设备支持传输特征数据;所述第一设备的特征提取能力用于指示所述第一设备支持提取所述第一媒体信息的第一特征数据;通过所述第二传输接口接收所述第一设备发送的能力协商响应消息;所述能力协商响应消息用于确认所述第一设备支持传输特征数据的传输协议。
一种可能的实现方式,所述接收来自第一设备的第一特征数据之前,还包括:
通过所述第二传输接口接收所述第一设备发送的能力协商请求消息;所述能力协商请求消息用于请求所述第二传输接口支持的传输协议,及所述第二设备的特征数据处理能力,所述第二设备的传输协议用于指示所述第二设备支持传输特征数据;所述第二设备的特征数据处理能力用于指示所述第二设备支持处理所述第一特征数据获得第一应用的结果的能力;通过所述第二传输接口向所述第一设备发送能力协商响应消息;所述能力协商响应消息用于确认所述第二设备支持传输特征数据的传输协议及所述第二设备的特征数据处理能力。
一种可能的实现方式,所述接收所述第一特征数据之前,还包括:
获取第一特征数据处理模型;所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果;第一特征提取模型的版本与所述第一特征数据处理模型的版本对应,所述第一特征提取模型用于对所述第一媒体信息进行特征提取。
通过上述方法,可以在第一设备和第二设备建立连接,并确定可以执行相应的协同处理的第一应用的任务后,第一设备和第二设备分别加载该第一应用对应的算法模型的输入部分(第一特征数据提取模型)和输出部分(第一特征数据处理模型)。从而实现第一设备和第二设备的第一应用协同。
一种可能的实现方式,所述能力协商响应消息还包括:所述第一设备中的特征提取模型的版本;或者,所述第二设备中的特征数据处理模型的版本。
通过上述方法,还可以通过能力协商响应,确认第一设备中的特征提取模型的版本和第二设备中的特征数据处理模型的版本是否可以完成第一应用的协同。
一种可能的实现方式,所述获取第一特征数据处理模型,包括:通过所述第二传输接口接收来自所述第一设备的第一特征数据处理模型,或者,接收来自服务器的第一特征数据处理模型,或者,读取所述第二设备存储的第一特征数据处理模型;
一种可能的实现方式,所述方法还包括:通过所述第二传输接口向第一设备发送第一特征提取模型;其中,所述第一特征提取模型的版本与所述第一特征数据处理模型的版本对应,所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果。一种可能的实现方式,所述方法还包括:获取第二特征数据处理模型,所述第二特征数据处理模型的版本与所述第二特征提取模型的版本对应,所述第二特征提取模型和所述第二特征数据处理模型为更新所述第一特征提取模型和所述第二特征数据处理模型后确定的。
通过上述方法,可以通过多种方式,使得第一设备获取第一特征提取模型,第二设备获取第一特征数据处理模型,实现更多灵活的AI应用协同方式。
一种可能的实现方式,所述方法还包括:接收第一训练特征数据;所述第一训练特征数据为所述第一设备根据所述第一特征提取模型对训练样本进行特征提取后确定的;根据所述第一训练特征数据,训练所述第一特征数据处理模型。
一种可能的实现方式,所述方法还包括:得到第一特征提取模型的反馈数据;所述反馈数据为所述第二设备根据所述第一训练特征数据训练后确定的;所述反馈数据用于所述第一设备训练所述第一特征提取模型;向所述第一设备发送所述反馈数据。
一种可能的实现方式,所述方法还包括:接收所述第二设备发送的第二特征数据;所述第二特征数据为所述第一设备根据采集的第二媒体信息及所述第二特征提取模型进行特征提取后确定的;根据所述第二特征数据处理模型对所述第二特征数据进行处理,得到所述第一应用的结果。
通过上述方法,可以利用第一设备和第二设备进行联合训练,合理利用第一设备和第二设备的算力,提升AI应用的性能。
一种可能的实现方式,所述方法还包括:通过所述第二传输接口向所述第一设备发送第一消息;所述第一消息用于指示所述第一设备采集媒体信息的状态。
通过上述方法,第二设备可以向第一设备发送第一消息,以调整第一设备采集媒体信息的状态,以更好的获得第一应用所需采集的媒体信息,提升第一应用的效果。
一种可能的实现方式,所述第一设备采集媒体信息的状态包括以下至少一项:开启 状态、关闭状态或采集媒体信息的参数。
一种可能的实现方式,所述方法还包括:通过所述第二传输接口向所述第一设备发送第二消息;所述第二消息用于指示所述第一设备获取第一数据;所述第一数据为以下一项:所述第一设备采集到的媒体信息,所述第一设备的参数,所述第一设备存储的数据,第一设备接收的数据。
通过上述方法,第二设备可以指示第一设备获取第一数据,使得在第一设备和第二设备之间传输特征数据的同时,兼容其他数据的传输,有助于提升第一设备和第二设备的实现AI应用功能所需的传输性能和传输场景的适应性。
一种可能的实现方式,所述方法还包括:通过所述第二传输接口接收来自所述第一设备的所述第一数据。
通过上述方法,第二传输接口可以支持多种数据的传输,提升传输性能和传输场景的适应性。
一种可能的实现方式,所述方法还包括:通过所述第二传输接口向第一设备发送第二消息;所述第二消息用于指示所述第一设备采集第三媒体信息的特征数据;通过所述第二传输接口接收来自所述第一设备发送的第三特征数据;所述第三特征数据为所述第一设备对采集的第三媒体信息进行特征提取后确定的。
通过上述方法,可以通过第二设备发送第二消息,以控制第一设备采集媒体信息,并传输相应的第三特征数据,例如,可以通过对第一应用的处理结果或AI应用的需要,确定需采集的第三媒体信息,从而灵活的调整第一设备采集的媒体信息,以使AI应用可以获得更好的结果,从而,整体提高AI应用的效果。
一种可能的实现方式,所述第一消息或所述第二消息为根据所述第一特征数据的处理结果确定的。
通过上述方法,第二设备可以基于第一设备传输的第一特征数据,确定第一应用的结果,并根据第一应用的结果生成第一消息或第二消息,从而向第一设备反馈相应的第一消息或第二消息,第一设备可以响应于第一消息或第二消息,调整媒体信息的采集、获取和传输,以使第一设备和第二设备更好的完成第一应用协同。
一种可能的实现方式,所述第一设备还包括显示单元;所述方法还包括:
响应于所述第一特征数据的处理结果,生成第三消息;所述第三消息用于指示所述第一设备显示的内容。
通过上述方法,可以通过第三消息,获得待显示的内容,该内容可以是第一应用的处理结果,也可以是其他第二设备需要第一设备显示的内容,从而,使得第一设备和第二设备更好的实现AI应用的协同,提升AI应用的使用体验。
一种可能的实现方式,所述第一设备的数量为N个;所述方法还包括:
通过所述第二传输接口接收第四消息;所述第四消息包括所述N个第一设备的M个第一特征数据;N、M为大于1的正整数;M大于或等于N;
根据所述M个第一特征数据对应的特征数据处理模型,对所述M个第一特征数据进行处理,得到所述第一应用的结果。
通过上述方法,可以实现传输多个第一设备中的M个第一特征数据,并在第二设备中对应的特征数据处理模型,处理M个第一特征数据,实现多个第一设备和第二设备之间的第一应用协同,提升第一应用协同的效果。
一种可能的实现方式,所述方法还包括:
通过所述第二传输接口向所述第一设备发送认证请求消息,所述认证请求消息用于请求所述第一设备是否与所述第二设备建立通信连接;所述通信连接用于确认所述第二设备控制所述第一设备的权限;通过所述第二传输接口接收所述第二设备发送的认证响应消息;所述认证响应消息用于确认所述第一设备是否与所述第二设备建立通信连接。
通过上述方法,可以通过第一设备和第二设备之间的认证,确认第二设备是否可以获得控制第一设备的权限,从而,在第二设备根据第一特征数据获得第一应用的结果后,调整第一设备采集媒体信息,有助于获取更好的第一应用的结果,提升AI应用的性能。
一种可能的实现方式,所述方法还包括:响应于所述第二设备发送的认证响应消息,为所述第一设备设置所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识;所述第一设备对应的设备标识及所述分布式系统的标识用于所述第一设备和所述第二设备进行通信;所述第一设备通过所述第一传输接口向所述第二设备发送认证成功消息;所述认证成功消息包括:所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识。
通过上述方法,还可以将第一设备和第二设备设置为分布式系统中的设备,以实现对第一设备和第二设备更好的管理,有利于利用多个设备实现AI应用协同。
一种可能的实现方式,所述第二设备包括第二模块;所述认证成功消息还包括以下至少一项:所述第二模块的标识,及所述第二模块在所述分布式系统中的标识。
通过上述方法,还可以将第一设备中的模块设置为分布式系统中的模块,从而,为第二设备控制各第一设备中的模块,协同完成AI应用做好准备。
一种可能的实现方式,所述第二设备还包括第三传输接口;所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第二设备发送的消息为通过所述第二传输接口封装为第二比特流数据后,通过所述第三传输接口发送的。
通过上述方法,通过第二传输接口封装数据,并通过第三传输接口向第一设备发送封装后的数据,从而,通过第三传输接口,可以兼容多种传输协议,也可以实现聚合传输等功能,从而提升第二设备的传输能力和传输媒体信息的兼容性。
一种可能的实现方式,所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第二设备接收的特征数据或消息为通过所述第三传输接口接收的第一比特流数据,并通过所述第二传输接口将所述第二比特流数据解封装后获得的。
通过上述方法,通过第二传输接口解封装第三传输接口接收的来自第一设备的数据,通过第三传输接口,可以兼容多种传输协议,也可以实现聚合传输等功能,从而提升第二设备的传输能力和传输媒体信息的兼容性。
第三方面,本申请提供一种电子设备,所述电子设备包括存储器和一个或多个处理器;其中,所述存储器用于存储计算机程序代码,所述计算机程序代码包括计算机指令;当所述计算机指令被所述处理器执行时,使得所述电子设备执行第一方面中任一项所述的方法。
第四方面,本申请提供一种电子设备,所述电子设备包括存储器和一个或多个处理器;其中,所述存储器用于存储计算机程序代码,所述计算机程序代码包括计算机指令;当所述计算机指令被所述处理器执行时,使得所述电子设备执行第一方面或第二方面中任一种可能实现方式中的方法。
第五方面,本申请提供一种媒体信息传输系统,包括:第三方面所述的电子设备和第 四方面所述的电子设备。
第六方面,本申请提供一种计算机可读存储介质,所述计算机可读存储介质包括程序指令,当所述程序指令在电子设备上运行时,使得所述电子设备执行如第一方面任一种可能的方法,或者,使得所述电子设备执行第二方面任一种可能的方法。
图1a为现有技术中的一种媒体信息发送设备的结构示意图;
图1b为现有技术中的一种媒体信息接收设备的结构示意图;
图1c为现有技术中的一种媒体信息接收设备的结构示意图;
图2a为本申请提供的一种媒体信息传输方法示意图;
图2b为本申请提供的一种AI算法模型的结构示意图;
图3a为本申请提供的一种第一设备的结构示意图;
图3b为本申请提供的一种第二设备的结构示意图;
图3c为本申请提供的一种分布式系统架构示意图;
图4a为本申请提供的一种AI应用协同的通信连接的建立方法的流程示意图;
图4b-图4c为本申请提供的一种第一设备的查找界面示意图;
图4d-图4e为本申请提供的一种AI应用协同的界面示意图;
图5a为本申请提供的一种分布式系统架构示意图;
图5b为本申请提供的一种媒体信息传输方法的流程示意图;
图5c为本申请提供的一种场景示意图;
图5d为本申请提供的一种AI应用的示意图;
图6a为本申请提供的一种分布式系统架构示意图;
图6b为本申请提供的一种媒体信息传输方法的流程示意图;
图6c为本申请提供的一种场景示意图;
图6d为本申请提供的一种AI应用的示意图;
图7a为本申请提供的一种分布式系统架构示意图;
图7b为本申请提供的一种媒体信息传输方法的流程示意图;
图7c为本申请提供的一种场景示意图;
图7d为本申请提供的一种AI应用的示意图;
图8为本申请实施例一种可能的电子设备的结构示意图;
图9为本申请实施例另一种可能的电子设备的结构示意图。
以下,先对本申请实施例中的部分用语进行解释说明,以便于本领域技术人员理解。
1)媒体信息
本申请涉及的媒体信息,可以包括:图像信息、音频信息、视频信息、传感器信息等通过第一设备采集的媒体信息。例如,采集的视频图像信息可以是音频、视频、可见光图像,或雷达、深度等信息。第一设备可以包括:摄像头、传感器、麦克风等具有媒体信息采集功能的设备或单元。
以媒体信息为图像信息为例,本申请实施例涉及的第一设备获取的媒体信息可以是原始图像,例如,可以是摄像头的输出图像,即摄像头将采集的物体反射的光信息转化为数字图像信号而得到的原始数据,该原始数据未经过加工处理。比如,原始图像可以是raw格式数据。该raw格式数据中可以包括物体的信息和摄像头参数。其中,摄像头参数可以包括感光度(international standardization organization,ISO)、快门速度、光圈值、白平衡等。原始图像也可以是ISP的输入图像,或者,原始图像也可以是网络神经单元,比如下文的神经网络处理器(neural-network processing unit,NPU)的输入图像。网络神经单元的输出图像可以是高动态范围图像(high dynamic range,HDR)图像,也可以是其他处理后的图像,在此不做限定。
本申请实施例涉及的第一设备获取的媒体信息,也可以是ISP的输出图像,由ISP对原始图像进行处理得到RGB格式或者YUV格式的图像,并且将RGB格式或者YUV格式的图像的亮度调整后得到的图像。其中,ISP将RGB格式或者YUV格式的图像的亮度调整的具体值,可以是用户设置的,也可以是手机在出厂时设置好的。第一设备获取的媒体信息也可以是处理器比如下文中第一设备的图形处理器(graphics processing unit,GPU)的输入图像。
需要说明的是,本申请实施例涉及的“媒体信息”,例如原始媒体信息、第一设备获取的媒体信息、第二设备处理后的媒体信息(例如,HDR图像)等,在媒体信息为图像时,可以是指图片,也可以是一些参数(比如,像素信息,颜色信息、亮度信息)的集合。
本申请实施例涉及的多个,是指大于或等于两个。
2)媒体信息的发送端设备
本申请实施例涉及的媒体信息的发送端设备,可以为具有媒体信息采集功能的设备。媒体信息可以包括图像信息、视频信息、音频信息、传感器信息中的一项或多项。以媒体信息为视频图像信息为例,媒体信息的发送端设备可以是具有视频图像采集功能的单元或设备,采集的视频图像信息可以是音频、视频、可见光图像,或雷达、深度等媒体信息。
媒体信息的发送端设备可以包括摄像头等视频采集单元,用于采集视频信息或图像信息,还可以包括麦克风等音频采集单元,用于采集音频信息。视频采集单元可以为光学镜头、图像传感器、麦克风等单元中的一项或多项,用于采集原始的媒体频信号(音频、图像或混合)。例如,媒体信息的发送端设备可以是:手机平板等移动终端、智能电视等智慧家庭终端、AR/VR头戴显示器、车载摄像头、外置摄像头等手机配件设备。例如,媒体信息的发送端设备可以是智慧屏等包括有媒体采集单元的终端设备。此时,媒体信息的发送端设备采集原始的音视频信息,经过处理后形成标准格式的音视频信号。媒体信息的发送端设备还可以作为媒体信息的发送端,经过压缩编码后通过传输接口或网络发送到接收端。其中,传输接口可以为HDMI、DP、USB等媒体传输接口。
在另一些可能的场景中,媒体信息的发送端设备还可以是获取媒体信息并发送该媒体信息的设备,例如,媒体信息的发送端设备可以是从网络或从本地存储单元获取媒体信息,并将该媒体信息发送给第二设备。此时,媒体信息的发送端设备可以不是具有该媒体信息采集功能的设备,即媒体信息的发送端设备可以仅为发送该媒体信息的发送功能的设备。
此时,媒体信息的发送端设备可以将获取的音视频媒体信息,经过压缩编码后通过传输接口或网络发送到接收端设备。其中,传输接口可以为HDMI、DP、USB等媒体传输接口。
如图1a所示,为现有技术中提供的一种媒体信息的发送端设备的结构示意图,在图1a所示的媒体信息的发送端设备中可以包括:媒体信号采集单元(例如,音频采集单元和视频采集单元),媒体编码单元(例如,音频编码器和视频编码器)和输出接口(例如,音频输出接口和视频输出接口)。其中,媒体信号获取单元可以有多种实现形式,举例来说,媒体信号获取单元,可以包括以下至少一项:媒体信号采集单元、媒体信号接收单元、存储单元。媒体信号采集单元,可以用于采集原始的媒体信号。例如,可以包括光学镜头、图像传感器、麦克风、雷达等媒体信号采集传感器或装置中的一项或多项。获取的媒体信息可以是音频、视频、可见光图像,或雷达、深度等信息。媒体信号接收单元,可以用于从网络或其他设备上接收媒体信号。存储单元,可以用于在发送端设备本地存储媒体信号,当然,还可以用于存储其他信息。媒体编码单元用于根据媒体编码协议、链路层和物理层协议对第一设备获取的媒体信号进行媒体编码和信道编码后得到物理层信号,将物理层信号传输到输出接口,从而发送给相应的媒体信息的接收端设备。可选的,媒体信息获取装置还可以包括媒体信号预处理单元。媒体信息预处理单元,可以用于对原始音视频媒体信号进行降噪、恢复等预处理。例如视频预处理单元,可以用于对原始的视频帧图像进行降噪、去马赛克等预处理。
3)媒体信息的接收端设备
媒体信息的接收端设备作为媒体信息的接收端,可以是媒体处理设备。接收端设备可以为手机、平板电脑、智能电视、车载计算机等终端设备。电子设备还可以称为终端设备。终端设备也可以称为用户设备、接入终端、用户单元、用户站、移动站、移动台、远方站、远程终端、移动设备、用户终端、终端、无线通信设备、用户代理或者用户装置。终端设备可以是手机(mobile phone)、平板电脑(pad)、带无线收发功能的电脑、虚拟现实(virtual reality,VR)终端、增强现实(augmented reality,AR)终端、工业控制(industrial control)中的无线终端、无人驾驶(self driving)中的无线终端、远程医疗(remote medical)中的无线终端、智能电网(smart grid)中的无线终端、运输安全(transportation safety)中的无线终端、智慧城市(smart city)中的无线终端、智慧家庭(smart home)中的无线终端等。蜂窝电话、无绳电话、个人数字助理(personal digital assistant,PDA),具有无线通信功能的手持设备、计算设备或者无线调整解调器的其它处理设备、车载设备、可穿戴设备等。
媒体信息的接收端设备还可以是机顶盒、分体扩展坞(DOCK)、智能电视、智能大屏、移动手机、平板电脑或者个人计算机(personal computer,PC)、智慧屏、手机、智能摄像头、智能音箱、耳机等终端设备。智慧屏可以是家庭中的影音娱乐中心,更是信息共享中心、控制管理中心、多设备交互中心。终端设备还可以是包含诸如个人数字助理和/或音乐播放器等功能的便携式终端,诸如手机、平板电脑、具备无线通讯功能的可穿戴设备(如智能手表)、车载设备等。便携式终端的示例性实施例包括但不限于搭载
或者其它操作系统的便携式终端。上述便携式终端也可以是诸如具有触敏表面(例如触控面板)的膝上型计算机(Laptop)等。还应当理解的是,在其他一些实施例中,上述终端也可以是具有触敏表面(例如触控面板)的台式计算机。
媒体信息的接收端设备也可以为机顶盒、显示屏、智能大屏、电视(television,TV)、移动手机或者其他具有媒体信息处理功能的终端设备中的处理器芯片,示例性地,该处理器芯片器可以是片上系统(system on chip,SoC)或基带芯片。第二设备还可以是部署有图形处理器(graphics processing unit,GPU)的计算设备、分布式计算设备等等。
如图1b所示,为本申请中提供的一种媒体信息的接收端设备的结构示意图,该媒体处理装置中可以包括:输入接口,媒体解码单元和媒体信息处理单元。其中,输入接口可以用于接收来自发送端(例如,媒体信息的发送端设备)发送的媒体信号。输入接口用于从传输信道接收物理层信号,媒体解码单元,用于根据链路层和物理层协议,从物理层信号中解码出媒体数据信号。
举例来说,媒体解码单元和媒体信息处理单元可以包括:解析器、音频解码器、视频解码器、视频后处理单元。其中每个单元可以通过硬件实现,也可以通过软件实现,或者可以通过硬件结合软件实现。例如,视频解码器、视频后处理单元等由硬件逻辑实现,媒体数据的AI分析、显示策略处理等单元可以由运行在硬件处理器上的软件代码来实现,音频解码器等其他单元可以通过软件实现。
一种可能的场景,媒体信息的接收端设备对从接口信道获取的编码后的信号,进行解压和解码后还原出音视频媒体信息,示例性的,mp4等格式的媒体文件经解析器解析后得到音频编码文件、视频编码文件等。其中,音频编码文件可以是音频基本码流(elementary stream,ES)数据,视频编码文件可以是视频ES数据。音频编码文件经音频解码器解码后得到音频数据;视频编码文件经视频解码器处理后得到视频帧。此外,媒体信息的接收端设备还可以用于将视频后处理得到的图像与音频数据同步,使得音频输出接口的输出和视频输出接口的输出同步,即使得音频输出接口输出的音频与视频输出接口输出的视频画面同步。
可选的,该媒体信息的接收端设备还可以包括显示单元,此时,可以将接收到的音视频媒体信息进行处理,并进行音视频媒体信息的播放。或者,显示单元也可以位于其他设备中,该设备为与媒体数据处理装置通过无线或有线建立通信连接的设备。例如,显示单元可以位于显示器(或者称为显示屏)、电视机、投影仪等终端设备中。该显示单元可以用于播放该数据处理装置处理后的媒体文件,还可以播放其他媒体文件。
另一种可能的场景,随着深度学习算法和NPU等计算硬件的发展,机器视觉、语音交互等人工智能(AI)应用快速普及,该数据处理装置可以对多媒体文件进行AI应用等处理,即该数据处理装置中还可以具备AI处理能力,用于实现相应的AI应用的功能。
在该场景下,在接收端从接口信道获取的信息集进行解压和解码后还原出音视频媒体信息之后,还可以执行图像处理等操作,利用接收端的AI软硬件处理系统执行后续的AI应用处理,获得媒体数据的AI分析结果。
举例来说,如图1c所示,该媒体信息的接收端设备还可以包括:神经网络训练单元和神经网络处理单元。
神经网络训练单元,用于利用标注数据训练AI算法模型。其中,AI算法模型的训练过程可以在服务器离线进行,也可以在设备端或云端在线进行。
神经网络处理单元,用于加载1个或多个AI算法模型,对媒体数据进行处理,获取AI算法模型的推理结果。例如,利用媒体信息的接收端设备的深度学习算法和NPU等AI软硬件系统对音视频信号进行AI处理,用于获取检测、分类、识别、定位、追踪等推理结果,该推理结果可以用于相应的AI应用的场景。例如,利用推理结果实现生物特征识别、环境识别、场景建模、机器视觉交互、语音交互等功能。
4)媒体信息的传输
本申请实施例中涉及的传输包括接收和/或发送。媒体信息的发送端和媒体信息的接收 端可以通过有线或无线等方式进行连接并传输媒体信息。传输接口的形态可以是有线传输的电信号、光纤传输的光信号、无线电信号、无线光信号等。媒体信息的发送端设备与媒体信息的接收端设备之间可以通过铜线、光纤等有线和/或无线通信协议建立物理上的信道连接。
如图1c所示,为本申请实施例的一种媒体信息传输的网络系统的架构示意图,所述网络系统中包括媒体信息的发送端设备和媒体信息的接收端设备。
以有线的方式为例,传输媒体信息的物理层信号可以通过传输信道进行传输。传输信道可以为铜线、光纤等物理传输信道。传输的媒体信息的信号可以为有线电信号、光信号等。其中,传输媒体信息的数据信号可以是HDMI协议的数据信号、DP协议的数据信号,或者是其他协议的数据信号。例如,电子设备用于传输媒体数据的接口标准包括:高清多媒体接口(high definition multimedia interface,HDMI)、USB接口、DP接口等。HDMI是一种传输无压缩数字高清多媒体(视频和/或音频)的接口。在数据传输上,HDMI使用最小化传输差分信号(transition minimized differential signaling,TMDS)技术。USB是一种串口总线标准,也是一种输入输出接口的技术规范。例如,USBType-C接口能够支持PD以及支持传输多媒体数据以外的其它数据。
为实现媒体信号的传输,一种可能的方式为,在进行媒体信息的传输之前,将媒体信息的发送端上获取的音视频媒体信息进行编码,经过音视频的编码后,再传输到媒体信息的接收端。此类方法会在信道中传输媒体信号,媒体信号在传输过程中的数据量较大,消耗计算资源多、成本高,系统整体延迟大,不利于对实时性要求高的AI应用。
为了能在较低的信道带宽下传输媒体信息,降低视频传输过程中的数据量,使得视频传输的时效性更高,另一种可能的方式,媒体信息的发送端可以对视频进行前处理(如分辨率压缩处理),即对媒体信息进行有损压缩编码,从而,将生成的有损压缩视频进行编码后发送给接收端设备。该方式下,由于进行了有损压缩,导致媒体信息的接收端无法完整获取到媒体信息,可能影响后续媒体信息的应用。另外,由于需要在发送端对媒体信息进行压缩编码,并需要在媒体信息的接收端进行恢复传输的媒体信息,编解码过程较为复杂,需要消耗更多的计算资源、导致系统整体延迟大,不利于应用于实时性要求高的AI应用中。
另外,由于在传输信道和接收端都是传输的是媒体信息,存在图像、音频等媒体内容隐私泄露的风险。即使经过加密处理媒体信号中的隐私信息也在传输过程和接收端存在泄漏的风险。
基于上述问题,如图2a所示,本申请提供一种媒体信息传输方法的流程示意图。在该方法中,根据AI应用(例如,第一应用)的需求,对应设置相应的算法模型。当通过第一设备(例如,媒体信息的发送端设备)和第二设备(例如,媒体信息的接收端设备)进行分布式的AI应用协同时,本申请中,可以将预先训练好的算法模型或在线训练获取的算法模型划分为两部分:输入部分和输出部分。
输入部分可以是第一特征提取模型。输出部分可以是对特征数据进行后处理的特征数据处理模型。第一特征提取模型可以为卷积神经网络(convolutional neural network)的特征提取单元,第一特征数据处理模型可以为任意的神经网络模型(如卷积神经网络模型),还可以为其他算法模型,在此不做限定。图2b中以卷积神经网络模型为例进行说明。
输入部分为特征提取部分(也可以称为特征提取模型),输入部分可以包含AI算法模 型的输入层,用于对获取的音视频等媒体信息进行特征提取,得到特征数据。例如,输出的特征数据可以是经过AI算法模型的输入层的卷积、加权等处理后的特征数据。图2b中的x1~x6为输入层中的卷积模块。输出部分为对特征数据进行后处理的部分(也可以称为特征数据处理模型),例如,输出部分包含模型的隐层和输出层,将经由图2b中的x1~x7为输入层中的卷积模块进行特征提取后的特征数据输入至隐层中的卷积模块,得到的数据输入到输出层,例如,输出层可以为分类器,得到AI算法模型的处理结果。
在通过第一设备和第二设备进行AI应用协同时,可以将AI应用对应的AI算法模型的输入部分加载到第一设备,将AI应用对应的AI算法模型的输出部分加载到第二设备。第一设备和第二设备都具备AI应用所需的软硬件能力,例如,包括NPU等处理AI算法模型的输入部分的计算硬件和用于AI应用协同的AI算法模型。以处理AI算法模型的处理器为NPU为例,此时,可以将输入部分部署到第一设备的NPU中,将输出部分部署到第二设备的NPU中。
步骤201:第一设备获取第一媒体信息。
其中,第一设备获取第一媒体信息的方式可以参考图1a中的媒体信息的发送端设备的获取方式,在此不再赘述。
步骤202:第一设备对所述第一媒体信息进行特征提取,确定所述第一媒体数据的第一特征数据。
其中,第一设备可以基于第一设备与第二设备进行AI应用协同中的AI应用,确定AI应用对应的AI算法模型的输入部分(第一特征提取模型),并在第一设备上加载该AI应用的AI算法模型的输入部分,对第一媒体信息进行特征提取。
步骤203:第一设备向第二设备发送所述第一特征数据。
步骤204:第二设备对第一特征数据进行处理,确定第一应用的结果。
将第一设备加载的AI算法模型的输入部分输出的特征数据,通过本申请中的传输接口传输到第二设备,并利用第二设备加载的AI算法模型的输出部分(第一特征数据处理模型)进一步处理后获得AI模型的最终输出结果。例如,第一设备获取的媒体信息可以不传输到第二设备,而是利用第一设备的NPU运行AI算法模型的输入部分,将媒体信息转换为特征数据后再传输,从而,通过支持传输特征数据的传输接口将特征数据传输到第二设备进行AI模型输出部分的处理。
通过上述方法,将发送媒体信息的第一设备中增加特征提取模型的处理能力,使得发送端无需发送媒体信息,而是发送将媒体信息进行特征提取后的特征数据。由于特征数据的数据量显著的低于原始的音视频信息,并且可以省去了额外的对媒体信息的压缩编码过程,降低系统功耗和延时,降低成本,还可以实现在较小的信道带宽条件下的实时传输,提升产品竞争力。另外,由于传输接口内传输的特征数据无法逆向转换为原始的媒体信息,实现更好的隐私保护能力。
在一些实施例中,在进行AI应用协同之前,AI算法模型的输入部分和输出部分可以分别存储在第一设备和第二设备,也可以都先存储在第一设备、第二设备或云端服务器,在此不做限定。
在第一设备和第二设备建立连接,并确定可以执行相应的协同处理的AI任务后,第一设备和第二设备分别加载该AI应用对应的算法模型的输入部分和输出部分,并确认第一设备和第二设备是否成功加载该AI应用算法模型的输入部分和输出部分。
以第二设备存储有算法模型的输入部分和输出部分为例。当第二设备确认第一设备加载AI应用对应的算法模型的输入部分失败时,则第二设备根据第一设备是否具备加载处理该AI应用的算法模型的输入部分的能力,确认第一设备是否可以执行该AI应用的协同。如果具备,则通过传输接口的数据传输信道将模型的输入部分的数据传输到第一设备,以使第一设备加载该AI算法模型的输入部分。
加载失败的原因可能有多种,一种可能的原因,例如,第一设备上并未存储有该AI应用对应的算法模型的输入部分。另一种可能的原因是,第一设备上存储的该AI应用对应的算法模型的输入部分的版本不是该AI应用所需的版本,此时,可以根据该AI应用对应的算法模型的模型标识及该AI应用对应的算法模型的输入部分的版本标识,从相应存储该AI应用对应的算法模型的输入部分的设备中获取。结合上述例子,可以是从第二设备中获取,也可以是从相应的服务器中获取,在此不做限定。
以网络中的服务器存储有算法模型的输入部分和输出部分为例。此时,当服务器确认第一设备加载AI应用对应的算法模型的输入部分失败时,则服务器根据第一设备是否具备加载处理该AI应用的算法模型的输入部分的能力,确认第一设备是否可以执行该AI应用的协同。如果具备,则通过传输接口的数据传输信道将模型的输入部分的数据传输到第一设备。相应的,当服务器确认第二设备加载AI应用对应的算法模型的输出部分失败时,则服务器根据第二设备是否具备加载处理该AI应用的算法模型的输出部分的能力,确认第二设备是否可以执行该AI应用的协同。如果具备,则通过传输接口的数据传输信道将模型的输出部分的数据传输到第二设备,以使第二设备加载该算法模型的输出部分。
需要说明的是,第一设备可以是多个,即多个第一设备中的第一特征提取模型作为第二设备的算法模型的输入部分,实现多个第一设备与第二设备的算法模型的协同AI应用。另外,针对不同的AI应用,第一设备中的第一特征提取模型也可以为相同的特征提取模型,也可以为不同的特征提取模型,可以基于具体的应用相应设置,在此不做限定。
根据AI应用的需求,确定对算法模型进行训练的标签信息。其中,标签信息可以是通过网络或人机交互输入人工标注后的标签信息,也可以是标准的标签信息、聚类信息等标签信息。从而,通过第二设备中的神经网络训练单元,利用确定的标签数据,训练AI算法模型。
AI算法模型的训练过程可以在服务器离线或在线进行训练的。也可以是在设备端或云端在线进行训练或优化的。例如,可以通过在线学习、迁移学习、增强学习和联邦学习等方法AI算法模型的参数进行更新,实现模型的再训练或再优化。可选的,模型也可以为利用本申请中的分布式系统协同训练或优化后获得的。本申请中,可以为每个训练后获得AI算法模型分配指定模型ID。
利用本申请的分布式系统进行AI算法模型的分布式协同训练,在训练模式下,由第一设备采集或输入模型的输入信息,传输接口可以传输第一设备的NPU,以输出的对训练样本进行特征提取的特征数据,在第二设备通过标签信息对第一设备发送的训练样本的特征数据进行训练,得到AI算法模型输出部分的反馈数据,第二设备可以将AI算法模型输出部分的反馈数据通过传输接口反向传输到模型的输入部分,用于对第一设备的输入部分进行训练,实现两部分网络参数的协同调整,从而实现第一设备与第二设备的协同训练。
以对第一特征提取模型和第一特征数据处理模型进行协同训练为例,此时,第一设备可以根据第一特征提取模型,对训练样本进行特征提取,生成第一训练特征数据;向所述 数据处理装置发送第一训练特征数据。第二设备根据接收到的第一训练特征数据及训练样本的标签信息,训练第一特征数据处理模型,确定训练第一特征提取模型的反馈数据。并将该反馈数据发送给第一设备。第一设备根据该反馈数据对第一特征提取模型进行训练。
在在线训练开始时,可以为离线训练获得的模型分配一个初始的版本ID(例如,初始的版本ID对应第一特征提取模型和第一特征数据处理模型);若对模型进行再训练或再优化后,更新一个指定的版本ID;当模型每次训练或在线优化后,再为模型的输入部分和输出部分分配一个更新后的版本ID(例如,更新后的版本ID对应第二特征提取模型和第二特征数据处理模型)。从而,通过模型ID和版本ID共同标识模型的输入部分和输出部分。进一步的,针对模型的输入部分和输出部分,还可以设置相应的标识。例如,针对模型的输入部分,可以设置模型ID和版本ID及输入部分ID。针对模型的输入部分,可以设置模型ID和版本ID及输出部分ID。另一种可能的实现方式,模型ID和版本ID可以针对模型的输入部分和输出部分相应设置。例如,模型ID包括模型输入部分ID和模型输出部分ID,版本ID包括模型版本的输入部分ID和模型版本的输出部分ID。具体实现方式本申请不做限定。
此时,第一设备在执行该AI算法模型对应的AI应用协同时,可以基于更新后的版本,即根据第二特征提取模型对采集的第二媒体信息进行特征提取,获得第二特征数据。从而,将第二特征数据发送给第二设备。此时,第二设备可以根据模型ID和更新后的版本ID,确定第二特征数据处理模型,从而,根据第二特征数据处理模型和第二特征数据,获得AI应用的AI算法模型的推理结果。
考虑到第一设备可以有多个的场景,此时,第二设备可以接收多个第一设备的训练数据对第二设备的神经网络模型进行训练,并相应生成多个第一设备中的模型训练的反馈数据,并将该多个反馈数据分别发送给多个第一设备,以便多个第一设备可以采用相应的反馈数据进行模型训练。
可选的,当AI应用协同的分布式系统中存在多个媒体信息采集单元、存储单元和媒体信号接收单元等多个单元时,可以为各个单元设置单元标识。从而,第二设备可以根据各个单元的单元ID进行寻址发送控制信息和用于训练AI算法模型的反馈数据。例如,在为媒体信息采集单元发送控制信息时,发送的控制信息的消息中可以携带有媒体信息采集单元的单元ID。在为神经网络训练单元发送反馈数据时,发送的控制信息的消息中可以携带有神经网络训练单元的单元ID。
如图3a所示,为本申请实施例提供的一种第一设备的结构示意图。第一设备200可以包括处理器210,外部存储器接口220,内部存储器221,传输接口230,媒体信息采集单元240,通信单元250。其中,媒体信息采集单元240可以包括:麦克风,耳机接口,音频单元,扬声器;传感器单元280,摄像头281等。通信单元250可以包括:天线1,天线2,移动通信单元251,无线通信单元252。
处理器210可以包括一个或多个处理单元,例如:处理器210可以包括应用处理器(application processor,AP),调制解调处理器,图形处理器(graphics processing unit,GPU),图像信号处理器(image signal processor,ISP),控制器,存储器,视频编解码器,数字信号处理器(digital signal processor,DSP),基带处理器,和神经网络处理单元,例如神经网络处理器(Neural-network Processing Unit,NPU)等。其中,不同的处理单元可以是独立的器件,也可以集成在一个或多个处理器中。其中,控制器可以是第一设备200的神经中枢和指挥 中心。控制器可以根据指令操作码和时序信号,产生操作控制信号,完成取指令和执行指令的控制。
本申请实施例中,神经网络处理器可以包括神经网络处理单元。神经网络处理单元,用于加载第一设备对应的AI算法模型的输入部分,对输入至神经网络处理单元的媒体信息(例如,第一设备获取的媒体信息)进行处理,输出该媒体信息的特征数据。
利用第一设备的神经网络处理单元上加载的AI算法模型的输入部分处理所述媒体信息,处理后获得抽象的特征数据;第一设备将特征信息经过第一设备的传输接口传输到第二设备,使得第二设备利用模型的输出部分进行进一步的处理并获得AI应用的输出结果。
其中,AI算法模型可以是服务器训练确定的,也可以是第一设备或第二设备单独训练确定的,还可以是第一设备与第二设备协同训练确定的。AI算法模型可以是离线训练的,也可以是在线训练的。基于第一设备使用AI算法模型的输入部分,第二设备使用AI算法模型的输出部分。因此,相应的,训练过程也可以是第一设备训练AI算法模型的输入部分,第二设备训练AI算法模型的输出部分,实现AI算法模型的协同训练。
可选的,神经网络处理器还可以包括:神经网络训练单元。该神经网络训练单元使得第一设备利用标注数据训练AI算法模型。该训练过程可以在第一设备离线进行,也可以在第一设备端在线进行,还可以与第二设备协同进行。
在协同进行训练时,可以将在线训练得到的训练数据发送给第二设备,以便第二设备可以根据第一设备得到的训练数据对第二设备的模型进行训练,并生成对第一设备中的模型训练的反馈数据,以便第一设备根据反馈数据对第一设备的神经网络模型进行训练。
处理器210中还可以设置存储器,用于存储指令和数据。在一些实施例中,处理器210中的存储器为高速缓冲存储器。该存储器可以保存处理器210刚用过或循环使用的指令或数据。如果处理器210需要再次使用该指令或数据,可从所述存储器中直接调用。避免了重复存取,减少了处理器210的等待时间,因而提高了系统的效率。
处理器210可以运行本申请实施例提供的媒体信息传输方法,以便于实现第一设备在AI应用下的多设备的协同,提升用户的体验。当处理器210运行本申请实施例提供的媒体信息传输方法后,处理器210可以根据获取的媒体信息,生成并发送特征数据。还可以接收来自第二设备的控制指令,用于控制第一设备的媒体采集单元,可选的,在第一设备包括有显示屏时,还可以接收来自第二设备的媒体信息及播放该媒体信息的指令时,可以通过显示屏播放该媒体信息。处理器210可以包括不同的器件,比如集成CPU和GPU时,CPU和GPU可以配合执行本申请实施例提供的媒体信息传输方法,比如媒体信息传输方法中部分算法由CPU执行,另一部分算法由GPU执行,以得到较快的处理效率。
内部存储器221可以用于存储计算机可执行程序代码,所述可执行程序代码包括指令。处理器210通过运行存储在内部存储器221的指令,从而执行第一设备200的各种功能应用以及数据处理。内部存储器221可以包括存储程序区和存储数据区。其中,存储程序区可存储操作系统,应用程序(比如相机应用,微信应用等)的代码等。存储数据区可存储第一设备200使用过程中所创建的数据(比如相机应用采集的图像、视频等)等。
内部存储器221还可以存储本申请实施例提供的数据传输算法对应的一个或多个计算机程序。该一个或多个计算机程序被存储在上述存储器221中并被配置为被该一个或多个处理器210执行,该一个或多个计算机程序包括指令,上述指令可以用于执行如图2a至图7b相应实施例中的各个步骤,该计算机程序可以用于实现本申请实施例中涉及的媒体信息 传输方法。当内部存储器221中存储的数据传输算法的代码被处理器210运行时,处理器210可以运行本申请实施例中涉及的媒体信息传输方法。
此外,内部存储器221可以包括高速随机存取存储器,还可以包括非易失性存储器,例如至少一个磁盘存储器件,闪存器件,通用闪存存储器(universal flash storage,UFS)等。
当然,本申请实施例提供的数据传输算法的代码还可以存储在外部存储器中。这种情况下,处理器210可以通过外部存储器接口220运行存储在外部存储器中的数据传输算法的代码,处理器210可以执行本申请实施例中涉及的媒体信息传输方法。
摄像头281(前置摄像头或者后置摄像头,或者一个摄像头既可作为前置摄像头,也可作为后置摄像头)用于捕获静态图像或视频。通常,摄像头281可以包括感光元件比如镜头组和图像传感器,其中,镜头组包括多个透镜(凸透镜或凹透镜),用于采集待拍摄物体反射的光信号,并将采集的光信号传递给图像传感器。图像传感器根据所述光信号生成待拍摄物体的原始图像。
第一设备200可以通过音频单元270,扬声器270A,受话器270B,麦克风270C,耳机接口270D,以及应用处理器等实现音频功能。例如音乐播放,录音等。可选的,第一设备200可以接收按键290输入,产生与第一设备200的用户设置以及功能控制有关的按键信号输入。
其中传感器单元280可以包括距离传感器,陀螺仪传感器,加速度传感器,接近光传感器,指纹传感器,触摸传感器,温度传感器,压力传感器、距离传感器、磁传感器、环境光传感器、气压传感器、骨传导传感器等,图中未示出。
第一设备200中的传输接口用于连接其他设备,以使第一设备200与其他设备进行媒体信息的传输。在一种可能的实现方式中,第一设备200中的传输接口可以包括第一传输接口和第三传输接口。通过第一传输接口与第一设备的处理器连接将待发送给第二设备300的数据封装为第一比特流数据,并通过第三传输接口发送给第二设备的第三传输接口。通过第三传输接口可以接收到来自第二设备发送的第二比特流数据,从而,通过第一传输接口解封装得到第二比特流数据对应的数据或消息(第二比特流数据为第二设备通过第二传输接口封装的数据或消息)。从而,使得通过第一设备的第三传输接口与第二设备的第三传输接口建立的传输信道可以支持双向传输。
在另一些实施例中,多个第一设备200还可以通过多个第一设备的第一传输接口封装发送给第二设备的第一特征数据。例如,N个第一设备需要发送M个第一特征数据,此时,可以通过N个第一设备中的第一传输接口封装M个第一特征数据为独立的比特流数据(可以是M个第一特征数据分装为M个比特流数据,也可以是根据N个第一设备分别分装为N个比特流数据),并通过第三传输接口将封装好的比特流数据打包为一个数据包(例如,第四消息)发送给第二设备的第三传输接口。从而,第二设备可以通过第三传输接口接收分别封装的N个第一设备的M个第一特征数据的第四消息,并通过第二传输接口将第四消息解封装,得到M个第一特征数据,并根据所述M个第一特征数据对应的特征数据处理模型,将M个第一特征数据转发至对应的特征数据处理模型中进行处理,得到第一应用的结果。
在该传输接口为有线的传输接口230时,适用于该传输接口230的线缆可以通过插入传输接口230,或从传输接口230拔出,实现和第一设备200的接触和分离。
在第一设备200的传输接口为无线通信接口时,可以通过天线1,天线2,移动通信单元251,无线通信单元252,调制解调处理器以及基带处理器等实现第一设备200的无线通信功能。
天线1和天线2用于发射和接收电磁波信号。第一设备200中的每个天线可用于覆盖单个或多个通信频带。不同的天线还可以复用,以提高天线的利用率。例如:可以将天线1复用为无线局域网的分集天线。在另外一些实施例中,天线可以和调谐开关结合使用。
移动通信单元251可以提供应用在第一设备200上的包括2G/3G/4G/5G等无线通信的解决方案。移动通信单元251可以包括至少一个滤波器,开关,功率放大器,低噪声放大器(low noise amplifier,LNA)等。移动通信单元251可以由天线1接收电磁波,并对接收的电磁波进行滤波,放大等处理,传送至调制解调处理器进行解调。移动通信单元251还可以对经调制解调处理器调制后的信号放大,经天线1转为电磁波辐射出去。在一些实施例中,移动通信单元251的至少部分功能单元可以被设置于处理器210中。在一些实施例中,移动通信单元251的至少部分功能单元可以与处理器210的至少部分单元被设置在同一个器件中。在本申请实施例中,移动通信单元251还可以用于与第二设备进行信息交互,即向第二设备发送媒体信息的传输请求,并将发送的媒体信息的传输请求封装成指定格式的消息,或者移动通信单元251可用于接收第二设备发送的媒体信息的传输请求。
调制解调处理器可以包括调制器和解调器。其中,调制器用于将待发送的低频基带信号调制成中高频信号。解调器用于将接收的电磁波信号解调为低频基带信号。随后解调器将解调得到的低频基带信号传送至基带处理器处理。低频基带信号经基带处理器处理后,被传递给应用处理器。
调制解调处理器还可以包括信道编码单元和解码单元,其中,信道编码单元用于根据链路层和物理层协议对第一设备获取的数据信号进行信道编码后得到物理层信号,将物理层信号传输到传输接口。其中,数据信号可以为通过对媒体信息进行特征提取后确定的特征数据,也可以是媒体信息的数据。当然,需要发送给第二设备的其他参数和设备状态等信息也可以通过传输接口根据相应的传输协议通过相应的传输信道传输至第二设备(例如,第二设备)。这些信息可以与媒体信息共同发送,也可以通过其他传输协议发送,具体实现方式本申请不做限定。还可以包括控制信息解码单元,用于解码第二设备发送的控制信号,还可以用于解码接收到的来自第二设备的反馈数据,该反馈信信息用于模型的在线训练优化。
一种可能的实现方式,应用处理器通过音频设备(不限于扬声器270A,受话器270B等)输出声音信号,或通过显示屏显示图像或视频。在一些实施例中,调制解调处理器可以是独立的器件。在另一些实施例中,调制解调处理器可以独立于处理器210,与移动通信单元251或其他功能单元设置在同一个器件中。
无线通信单元252可以提供应用在第一设备200上的包括无线局域网(wireless local area networks,WLAN)(如无线保真(wireless fidelity,Wi-Fi)网络),蓝牙(bluetooth,BT),全球导航卫星系统(global navigation satellite system,GNSS),调频(frequency modulation,FM),近距离无线通信技术(near field communication,NFC),红外技术(infrared,IR)等无线通信的解决方案。无线通信单元252可以是集成至少一个通信处理单元的一个或多个器件。无线通信单元252经由天线2接收电磁波,将电磁波信号调频以及滤波处理,将处理后的信号发送到处理器210。无线通信单元252还可以从处理器210接收待发送的信号, 对其进行调频,放大,经天线2转为电磁波辐射出去。本申请实施例中,无线通信单元252,用于与第二设备建立连接,通过第二设备协同完成AI应用的任务。或者无线通信单元252可以用于接入接入点设备,向第二设备发送特征数据的媒体信息的传输请求对应的消息,或者接收来自第二设备发送的媒体信息的传输请求对应的消息。可选地,无线通信单元252还可以用于接收来自其他设备的媒体信息。
如图3b所示,为本申请实施例提供的一种第二设备的结构示意图。
第二设备300可以包括处理器310,外部存储器接口320,内部存储器321,传输接口330,天线11,天线12,移动通信单元351,无线通信单元352。
处理器310可以包括一个或多个处理单元,例如:处理器310可以包括应用处理器(application processor,AP),调制解调处理器,图形处理器(graphics processing unit,GPU),图像信号处理器(image signal processor,ISP),控制器,存储器,视频编解码器,数字信号处理器(digital signal processor,DSP),基带处理器,和神经网络处理器(Neural-network Processing Unit,NPU)等。其中,不同的处理单元可以是独立的器件,也可以集成在一个或多个处理器中。其中,控制器可以是第二设备300的神经中枢和指挥中心。控制器可以根据指令操作码和时序信号,产生操作控制信号,完成取指令和执行指令的控制。
本申请实施例中,神经网络处理器可以包括神经网络处理单元。神经网络处理单元,用于加载装载1个或多个AI算法模型的输出部分。根据特征数据对应的AI算法模型的输出部分,对特征数据进行处理,将解码后的特征数据传输到第二设备的神经网络处理单元,利用神经网络单元的加载的AI算法模型的输出部分对特征数据进行处理,获得AI算法处理的最终的推理结果;获取检测、分类、识别、定位、追踪等推理结果。将推理结果输出给人工智能应用,人工智能应用利用推理结果实现生物特征识别、环境识别、场景建模、机器视觉交互、语音交互等功能。
可选的,本申请实施例中的第二设备还可以包括:神经网络训练单元,该神经网络训练单元使得第二设备利用标注数据训练AI算法模型。该训练过程可以在第二设备离线进行,也可以在第二设备端在线进行,还可以与第一设备协同进行。可选的,在第一设备和第二设备协同对AI模型进行联合的在线训练和优化时,可以将在线训练得到的反馈数据可以通过接口系统从模型输出部分经过传输接口发送给模型输入部分,以便第一设备根据反馈数据对第一设备的神经网络模型进行训练。此时,第二设备还可以将AI模型的在线训练反馈数据发送给第一设备。
考虑到第一设备可以有多个的场景,此时,第二设备可以接收多个第一设备的训练数据对第二设备的神经网络模型进行训练,并相应生成多个第一设备中的模型训练的反馈数据,并将该多个反馈数据分别发送给多个第一设备,以便多个第一设备可以采用相应的反馈数据进行模型训练。
处理器310中还可以设置存储器,用于存储指令和数据。在一些实施例中,处理器310中的存储器为高速缓冲存储器。该存储器可以保存处理器310刚用过或循环使用的指令或数据。如果处理器310需要再次使用该指令或数据,可从所述存储器中直接调用。避免了重复存取,减少了处理器310的等待时间,因而提高了系统的效率。
处理器310可以运行本申请实施例提供的媒体信息传输方法,以便于实现第二设备在AI应用下与第一设备的协同,提升用户的体验。当处理器310运行本申请实施例提供的媒 体信息传输方法后,处理器310可以根据获取的媒体信息,生成并发送特征数据。还可以向第一设备发送控制指令,用于控制第一设备的媒体采集单元,可选的,在第一设备包括有显示屏时,还可以向第一设备发送媒体信息及播放该媒体信息的指令时,可以通过显示屏播放该媒体信息。可选的,在第二设备包括有显示屏时,还可以接收第一设备发送的媒体信息,或者接收到来自第一设备的特征数据后,对特征数据进行处理,获得的AI算法处理的推理结果为待播放的媒体信息时,可以通过显示屏播放该媒体信息。
处理器310可以包括不同的器件,比如集成CPU和GPU时,CPU和GPU可以配合执行本申请实施例提供的媒体信息传输方法,比如媒体信息传输方法中部分算法由CPU执行,另一部分算法由GPU执行,以得到较快的处理效率。
内部存储器321可以用于存储计算机可执行程序代码,所述可执行程序代码包括指令。处理器310通过运行存储在内部存储器321的指令,从而执行第二设备300的各种功能应用以及数据处理。内部存储器321可以包括存储程序区和存储数据区。其中,存储程序区可存储操作系统,应用程序(比如相机应用,微信应用等)的代码等。存储数据区可存储第二设备300使用过程中所创建的数据(比如相机应用采集的图像、视频等)等。
内部存储器321还可以存储本申请实施例提供的数据传输算法对应的一个或多个计算机程序。该一个或多个计算机程序被存储在上述存储器321中并被配置为被该一个或多个处理器310执行,该一个或多个计算机程序包括指令,上述指令可以用于执行如图2a至图7a相应实施例中的各个步骤,该计算机程序可以用于实现本申请实施例中涉及的媒体信息传输方法。当内部存储器321中存储的数据传输算法的代码被处理器310运行时,处理器310可以运行本申请实施例中涉及的媒体信息传输方法。
此外,内部存储器321可以包括高速随机存取存储器,还可以包括非易失性存储器,例如至少一个磁盘存储器件,闪存器件,通用闪存存储器(universal flash storage,UFS)等。
当然,本申请实施例提供的数据传输算法的代码还可以存储在外部存储器中。这种情况下,处理器310可以通过外部存储器接口320运行存储在外部存储器中的数据传输算法的代码,处理器310可以执行本申请实施例中涉及的媒体信息传输方法。
第二设备300中的传输接口330用于连接其他设备,以使第二设备300与其他设备进行媒体信息的传输。
在一种可能的实现方式中,第二设备300中的传输接口可以包括第二传输接口和第三传输接口。第二传输接口与第二设备的处理器连接,通过第二传输接口将待发送给第二设备300的数据封装为第二比特流数据,并通过第三传输接口发送给第一设备的第三传输接口。通过第三传输接口可以接收到来自第一设备的第一比特流数据,并通过第二传输接口解封装得到第一设备发送的特征数据、数据和、控制信息、反馈数据、握手数据、消息等。从而,使得通过第一设备的第三传输接口与第二设备的第三传输接口建立的传输信道可以支持双向传输。
在另一些实施例中,第二设备300还可以通过第二设备的第二传输接口接收第四消息;所述第四消息包括所述N个第一设备的M个第一特征数据;N、M为大于1的正整数;M大于或等于N;具体的,可以通过第三传输接口接收分别封装的N个第一设备的M个第一特征数据的第四消息(第四消息中可以是封装为N个数据包,也可以是封装为M个数据包,在此不做限定),并通过第二传输接口将第四消息解封装,得到M个第一特征数据,并根据所述M个第一特征数据对应的特征数据处理模型,将M个第一特征数据转发至对 应的特征数据处理模型中进行处理,得到第一应用的结果。
在该传输接口330为有线的传输接口时,适用于该传输接口的线缆可以通过插入传输接口330,或从传输接口330拔出,实现和第二设备300的接触和分离。
在第二设备300的传输接口为无线通信接口时,第二设备300的无线通信功能可以通过天线11,天线12,移动通信单元351,无线通信单元352,调制解调处理器以及基带处理器等实现。
天线11和天线12用于发射和接收电磁波信号。第二设备300中的每个天线可用于覆盖单个或多个通信频带。不同的天线还可以复用,以提高天线的利用率。例如:可以将天线11复用为无线局域网的分集天线。在另外一些实施例中,天线可以和调谐开关结合使用。
移动通信单元351可以提供应用在第二设备300上的包括2G/3G/4G/5G等无线通信的解决方案。移动通信单元351可以包括至少一个滤波器,开关,功率放大器,低噪声放大器(low noise amplifier,LNA)等。移动通信单元351可以由天线11接收电磁波,并对接收的电磁波进行滤波,放大等处理,传送至调制解调处理器进行解调。移动通信单元351还可以对经调制解调处理器调制后的信号放大,经天线11转为电磁波辐射出去。在一些实施例中,移动通信单元351的至少部分功能单元可以被设置于处理器310中。在一些实施例中,移动通信单元351的至少部分功能单元可以与处理器310的至少部分单元被设置在同一个器件中。在本申请实施例中,移动通信单元351还可以用于与第二设备进行信息交互,即向第一设备发送媒体信息的传输请求,其接收的媒体信息的传输请求可以封装成指定格式的消息,或者移动通信单元351可用于向第一设备发送的媒体信息的传输指令或向第一设备发送控制信息的消息。
调制解调处理器可以包括调制器和解调器。其中,调制器用于将待发送的低频基带信号调制成中高频信号。解调器用于将接收的电磁波信号解调为低频基带信号。随后解调器将解调得到的低频基带信号传送至基带处理器处理。低频基带信号经基带处理器处理后,被传递给应用处理器。
调制解调处理器还可以包括信道编码单元和解码单元。
其中,信道解码单元可以根据链路层和物理层协议,从接收到的第一设备的物理层信号中解码出第一设备发送的数据信号。数据信号可以为第一设备通过对媒体信息进行特征提取后确定的特征数据(该特征数据可以作为神经网络处理单元的输入),也可以是媒体信息的数据。当然,需要第二设备接收的其他参数和设备状态等信息也可以通过传输接口根据相应的传输协议通过相应的传输信道传输。这些信息可以与媒体信息共同发送,也可以通过其他传输协议发送,具体实现方式本申请不做限定。
此时,信道编码单元可以用于对第二设备发送的数据信号进行编码。例如,数据信号可以为向第一设备发送的控制指令。控制指令可以经过第二设备的信道编码单元按照接口传输协议进行信道编码;编码后的控制信号经过传输接口调制后发送到控制信道中,通过第二设备的传输接口和控制信道传输到第一设备,使得第一设备可以通过该控制信道接收到该控制指令。信道编码单元还可以用于编码向第一设备发送的反馈数据,该反馈信信息用于第一设备的模型的在线训练优化。
一种可能的实现方式,应用处理器通过音频设备输出声音信号,或通过显示屏显示图像或视频。在一些实施例中,调制解调处理器可以是独立的器件。在另一些实施例中,调制解调处理器可以独立于处理器310,与移动通信单元351或其他功能单元设置在同一个 器件中。
无线通信单元352可以提供应用在第二设备300上的包括无线局域网(wireless local area networks,WLAN)(如无线保真(wireless fidelity,Wi-Fi)网络),蓝牙(bluetooth,BT),全球导航卫星系统(global navigation satellite system,GNSS),调频(frequency modulation,FM),近距离无线通信技术(near field communication,NFC),红外技术(infrared,IR)等无线通信的解决方案。无线通信单元352可以是集成至少一个通信处理单元的一个或多个器件。无线通信单元352经由天线12接收电磁波,将电磁波信号调频以及滤波处理,将处理后的信号发送到处理器310。无线通信单元352还可以从处理器310接收待发送的信号,对其进行调频,放大,经天线12转为电磁波辐射出去。本申请实施例中,无线通信单元352,用于与第一设备建立连接,通过与第一设备协同完成AI应用的任务。在一些实施例中,无线通信单元352还可以用于接入接入点设备,接收第一设备发送的特征数据的传输请求对应的消息,或者向第一设备发送媒体信息的传输请求对应的消息,向第一设备发送控制信息对应的消息。可选地,无线通信单元352还可以用于接收来自其他第一设备的媒体信息或其他设备的信息。
如图3c所示,为本申请实施例提供的一种第一设备和第二设备组成的AI应用协同的分布式系统的架构示意图。其中,本申请实施例提供的传输接口,可以为第一设备中的传输接口230或第一设备的通信单元250,和第二设备中的传输接口330或第二设备的通信单元350),图中以第一设备中的传输接口230和第二设备中的传输接口330为例进行说明。本申请实施例提供的传输接口可以适用于多种传输协议,也可以称为聚合接口,或者称为新接口(NEW口),也可以采用其它的名字,本申请实施例对此不作限定。
本申请实施例中的传输接口协议支持传输第一设备输出的特征数据。AI模型的输入部分处理后得到的特征数据,其数据量远低于原始的音视频媒体数据,因此其占用较低的带宽,解决了带宽传输高,消耗的带宽资源较多的问题。尤其是在WIFI、蓝牙等带宽较小的无线通信技术下,实现低带宽的传输接口实时传输,为实现实时的分布式AI处理创造条件。另外,由于通过所传输的特征数据无法恢复出原始的媒体信息,可解决潜在的隐私数据泄露风险,提高了媒体信息传输的数据安全性。
在一些实施例中,本申请实施例中的传输接口协议可以支持第一设备和第二设备对NPU架构、加载的AI算法模型的模型标识(ID)和版本标识(ID)等AI处理的软硬件能力进行协商,确定是否可以协调组成AI应用协同的分布式系统,以完成相应的AI应用的处理任务;在能力协商过程中,传输接口支持AI算法模型的双向传输,例如,可以将第一设备或第二设备存储的AI算法模型或网络获取的AI算法模型中的部分或全部传输到第一设备或第二设备上,用于实现AI算法模型的输入部分和输出部分的加载。
可选的,通过本申请实施例中的传输接口,可以将第二设备输出的在线训练的反馈数据发送给第一设备,用于对第一设备上的AI算法模型的输入部分信息在线训练或在线优化,并将进一步训练的特征数据反馈给第二设备,从而实现第一设备与第二设备的AI算法模型的在线训练和在线优化。
在另一些实施例中,传输接口还可以用于传输多种类型的数据。传输接口也可以传输AI模型输入部分输出特征数据,还可以双向传递握手信号、控制信号等控制消息。传输接口也可以传输媒体信息的信号或者其他数据信号。传输接口可以对上述信号进行同时的聚合传输和双向传输。例如,传输接口可支持媒体信息和AI特征数据的兼容传输,传输接 口可以传输标准的媒体信息数据,也具备传输经过采集端NPU处理后特征数据。传输接口也可以传输媒体信号、其他数据信号,传输接口也可以传输握手和控制信号。
在一些实施例中,传输接口可以为单向传输接口,也可以为双向传输接口。以单向传输接口为例,在发送端设置发送接口,在接收端设置接收接口。从而实现发送端向接收端传输媒体数据的功能。一种示例中,传输接口可以为双向传输接口,此时,传输接口具备发送功能也具备接收功能,即支持双向数据传输。比如传输接口支持发送和接收数据信号,也即,传输接口既可以作为数据信号的发送端,也可以作为数据信号的接收端。
在一些实施例中,传输接口具有数据聚合传输能力,例如,在带宽允许的条件下,接口的协议可以数据打包和混合等技术支持媒体信息和AI特征数据的在同一信道中同时传输。
在一些实施例中,传输接口可以传输原始的或压缩的媒体信息,例如,传输接口可以通过配置多个信道,支持媒体信息和AI特征数据的双向传输。
第一设备200中的传输接口可以包括第一传输接口和第三传输接口。通过第一传输接口与第一设备的处理器连接将待发送给第二设备300的数据封装为第一比特流数据,并通过第三传输接口发送给第二设备的第三传输接口。通过第三传输接口可以接收到来自第二设备发送的第二比特流数据,从而,通过第一传输接口解封装得到第二比特流数据对应的数据或消息(第二比特流数据为第二设备通过第二传输接口封装的数据或消息)。
第二设备300中的传输接口可以包括第二传输接口和第三传输接口。第二传输接口与第二设备的处理器连接,通过第二传输接口将待发送给第二设备300的数据封装为第二比特流数据,并通过第三传输接口发送给第一设备的第三传输接口。通过第三传输接口可以接收到来自第一设备的第一比特流数据,并通过第二传输接口解封装得到第一设备发送的特征数据、数据和、控制信息、反馈数据、握手数据、消息等。从而,使得通过第一设备的第三传输接口与第二设备的第三传输接口建立的传输信道可以支持双向传输。
在另一些实施例中,多个第一设备200中的第三传输接口为一个第三传输接口,此时,由多个第一设备中的第一传输接口封装后的多个第一比特流数据,可以通过该第三传输接口统一发送给第二设备300的第三传输接口。例如,N个第一设备生成M个第一特征数据,并通过N个第一设备的N个第一传输接口对M个第一特征数据分别封装,并通过第三传输接口打包为第四消息。通过第二设备300的第三传输接口接收该第四消息,从而,通过第二传输接口解封装M个第一特征数据,并根据所述M个第一特征数据对应的特征数据处理模型,将所述M个第一特征数据转发至对应的特征数据处理模型中进行处理。
需要说明的是,此处的数据信号可以为多媒体数据,也可以为本申请实施例中涉及的特征数据,还可以是用于建立传输链路的控制信息、还可以是用于传输其他参数和其他数据信号,在此不做限定。
下面举例说明本申请中传输接口具体传输过程。
在一种可能的实现方式中,发送端对传输的数据进行压缩和加密;将传输的数据经过信道编码传输至传输接口,再经调制后传输到接口的物理层信道。
以第一设备为向第二设备发送特征数据的发送端为例。第一设备还可以包括信道编码单元,通过信道编码单元对所述特征数据按照传输接口或标准约定的数据传输协议进行信道编码得到编码信号。可选的,第一设备在对特征数据进行信道编码前,还可以对特征数据进行数据压缩,以进一步减少数据传输量。可选的,第一设备在对特征数据进行信道编 码,还可以对信道编码后的特征数据进行加密,按照传输接口系统或标准约定的电气层和物理层传输协议将信道编码后的特征数据调制为物理层信号,从输出接口发送到的传输信道。
在一种可能的实现方式中,接收端接口将物理层信号解调后,可以进行信道解码后获得传输的数据;相应的,接收端还可以对解码后的信号进行解压缩和解密。
以第二设备为接收端接收第一设备发送的特征数据为例。第二设备还可以包括信道解码单元,用于接收第一设备的特征数据。例如,可以通过第二设备的输入接口,从传输信道中接收物理层信号,对物理层信号进行解调获取编码信号。通过信道解码单元,对接收的编码信号按照传输接口的协议进行信道解码,获得第二设备发送的特征数据。可选的,在该编码信号为加密和/或压缩后的信号时,该信道解码单元还可以对该编码信号进行解密和解压。
可选的,第一设备200或第二设备300还可以包括:显示屏,用于显示图像,视频等。显示屏包括显示面板。显示面板可以采用液晶显示屏(liquid crystal display,LCD),有机发光二极管(organic light-emitting diode,OLED),有源矩阵有机发光二极体或主动矩阵有机发光二极体(active-matrix organic light emitting diode的,AMOLED),柔性发光二极管(flex light-emitting diode,FLED),Miniled,MicroLed,Micro-oLed,量子点发光二极管(quantum dot light emitting diodes,QLED)等。在一些实施例中,第一设备200或第二设备300可以包括1个或S个显示屏,S为大于1的正整数。显示屏可用于显示由用户输入的信息或提供给用户的信息(例如,视频信息,语音信息,图像信息,文字信息等)以及各种图形用户界面(graphical user interface,GUI)。例如,显示屏可以显示照片、视频、网页、或者文件等。可选的,显示屏可以显示如图4b所示的图形用户界面。其中,如图4b所示的图形用户界面上可以包括状态栏、可隐藏的导航栏、时间和天气小组件(widget)、以及应用的图标,例如浏览器图标等。状态栏中包括运营商名称(例如中国移动)、移动网络(例如4G)、时间和剩余电量。导航栏中包括后退(back)键图标、主屏幕(home)键图标和前进键图标。此外,可以理解的是,在一些实施例中,状态栏中还可以包括蓝牙图标、Wi-Fi图标、外接设备图标等。还可以理解的是,在另一些实施例中,图4b所示的图形用户界面中还可以包括Dock栏,Dock栏中可以包括常用的应用图标等。当处理器210检测到用户的手指(或触控笔等)针对某一应用图标的触摸或手势事件后,响应于该触摸或手势事件,打开与该应用图标对应的应用的用户界面,并在显示屏上显示该应用的用户界面。示例性的,第一设备200或第二设备300的显示屏显示主界面,主界面中包括多个应用(比如相机应用、微信应用等)的图标。用户通过触摸传感器点击主界面中相机应用的图标,触发处理器210启动相机应用,打开摄像头。显示屏显示相机应用的界面,例如取景界面。可选的,第一设备200或第二设备300可以利用马达产生振动提示(比如来电振动提示)。第一设备200或第二设备300中的指示器可以是指示灯,可以用于指示充电状态,电量变化,也可以用于指示消息,未接来电,通知等。
可选的,第一设备200或第二设备300可以通过音频单元,以及应用处理器等实现音频功能。例如音乐播放,录音等。可选的,音频单元可以包括:扬声器,受话器,麦克风,耳机接口中的一项或多项。可选的,第一设备200或第二设备300可以接收按键输入,产生与第一设备200或第二设备300的用户设置以及功能控制有关的按键信号输入。
在本申请实施例中,显示屏可以是一个一体的柔性显示屏,也可以采用两个刚性屏以 及位于两个刚性屏之间的一个柔性屏组成的拼接显示屏。
应理解,在实际应用中,第一设备200可以包括比图3a所示的更多或更少的部件,第二设备300可以包括比图3b所示的更多或更少的部件,本申请实施例不作限定。图示第一设备200或第二设备300仅是一个范例,并且第一设备200或第二设备300可以具有比图中所示出的更多的或者更少的部件,可以组合两个或更多的部件,或者可以具有不同的部件配置。图中所示出的各种部件可以在包括一个或多个信号处理和/或专用集成电路在内的硬件、软件、或硬件和软件的组合中实现。
如图4a所示,为本申请实施例的一种媒体信息传输方法的流程示意图,第一设备与第二设备建立AI应用协同的通信连接,通过第一设备和第二设备之间互相配合,实现第一设备和第二设备协同完成一个AI任务。下面以第二设备具有显示屏,且第二设备发起建立AI应用协同的分布式系统的通信连接方式为例进行说明。在一些实施例中,还可以是第一设备主动发起建立与第二设备的通信连接,可以参考该实施例的方式实施,在此不做限定。具体可以包括:
步骤401:第二设备通过设备发现协议,发现第一设备。
第一设备与第二设备可以通过各自的传输接口及相应的有线或无线信道实现相互连接,例如,可以通过蓝牙、NFC、WIFI建立无线通信连接,还可以通过有线的方式建立通信连接。
步骤402:第二设备向第一设备发送能力协商请求消息。
其中,能力协商请求消息用于请求所述第一设备支持的传输协议,所述第一设备的传输协议用于指示所述第一设备是否支持传输特征数据。
第二设备可以根据本申请实施例中适应的设备发现协议,发现第一设备。进而,通过传输接口中控制信道和握手协议进行第一设备与第二设备的能力协商,使得第二设备确定第一设备的类型,支持AI处理的软件能力,支持AI处理的硬件能力,支持的媒体信息的传输协议等信息,从而确定第一设备是否可以与第二设备建立通信连接,用于实现相应的AI应用协同的功能。
在一些实施例中,可以是通过相应的AI应用触发建立第一设备与第二设备组成的分布式系统,还可以通过其他方式触发建立第一设备与第二设备组成的分布式系统,也可以是用户主动发起建立第一设备与第二设备的通信连接,以建立的第一设备与第二设备组成的分布式系统。例如,用户可以通过设置有AI应用协同功能的界面的第一设备或第二设备相应设置第一设备或第二设备的AI应用协同功能。例如,可以在第一设备上设置对应第一设备的AI应用协同功能的界面,使得用户可以在该界面上对第一设备的AI应用协同功能进行设置。也可以在第二设备上设置对应第二设备的AI应用协同功能的界面,从而,用户可以在该界面上对第一设备的AI应用协同功能进行设置。
考虑到一些场景中,第一设备或第二设备没有显示界面的功能,此时,可以在具有显示界面功能的设备上,对第一设备和第二设备的AI应用协同功能进行设置,即将第一设备和第二设备作为一个整体进行AI应用协同功能设置。例如,第一设备具有显示界面的功能时,可以对第一设备的AI应用协同功能进行设置,还可以对第二设备的AI应用协同功能进行设置,从而确定第一设备与第二设备组成的AI应用协同功能进行设置,在此不做限定。
下面以用户主动触发的方式为例进行说明。
例如,第二设备可以设置有AI应用协同功能的界面。示例性的,如图4b所示,为第二设备的AI应用协同功能的控制界面410,用户可以对该控制界面410进行操作,以设置是否与第一设备建立AI应用协同功能。例如,用户可以基于AI应用协同功能的开启/关闭控件420,开启/关闭AI应用协同功能。
用户打开第二设备的AI应用协同功能。在第二设备的AI应用协同功能的界面中可以显示第二设备发现的第一设备的标识。第一设备的标识可以包括第一设备的设备图标和/或第一设备作为AI应用协同的特征数据发送端时的发现名。
在一些实施例中,响应于用户的点击操作,第二设备显示第一查找设备界面430,该第一查找设备界面430可以为第二设备的AI应用协同服务的查找设备界面,该查找设备界面430中可以包括可发现的第一设备的标识,具体的,如图4c所示,查找设备界面中包括第一设备的图标450和该第一设备作为AI应用协同的特征数据发送端时的发现名440。这样,能够方便用户区分第二设备发现的AI应用协同的特征数据发送端的设备的类型,如AI应用协同的特征数据发送端的设备类型是分体式智慧屏,手机配件,还是监控设备,车载传感器设备等。如图4c所示,第一设备可以是智慧屏1,相机1,耳机1,AR眼镜。在本申请其他一些实施例中,第一查找设备界面中也可以不按照协同服务端的设备的类型对发现的设备进行区分。
示例性的,在其他一些实施例中,用户可以对第二设备的通知栏或任务栏中的提示框进行操作,响应于上述操作,第二设备打开上述第一查找设备界面。在另外一些实施例中,用户可以对第二设备的通知栏或任务栏中的相关图标进行操作,响应于上述操作,第二设备打开上述第一查找设备界面。响应于对第一设备的选择操作,第二设备可以向第一设备发送能力协商请求。
其中,该能力协商请求可以为通过第二设备的传输接口中的控制信道和相应的握手协议向第一设备发送的握手请求消息。该能力协商请求用于请求第一设备的媒体信息获取能力,媒体信息的主要参数,AI软硬件处理能力,传输接口支持的传输协议等能力信息。
步骤403:第一设备向第二设备发送能力协商响应消息。
其中,所述能力协商响应消息可以用于确认第一设备支持传输特征数据的传输协议。可选的,该能力协商响应消息还可以用于确认第一设备的能力信息。例如,所述能力协商响应消息还包括以下至少一项:所述第一设备的特征提取能力,及所述第一设备中的特征提取模型的版本。
在一些实施例中,第一设备在接收到第二设备发送的能力协商请求消息后,可以向第二设备返回能力协商响应,该能力协商响应消息中可以包括:第一设备的媒体信息获取能力,媒体信息的主要参数,AI软硬件处理能力,传输接口支持的传输协议等能力信息。
响应于第一设备的能力协商响应消息,第一设备可以根据能力协商响应消息,确定第一设备是否能够通过各自的传输接口,组成一个分布式系统,用于AI应用协同(例如,第一应用的协同)。即第一设备的传输接口和第二设备的传输接口是否可以支持传输媒体信息的特征数据,以及第一设备是否具有对媒体信息进行特征提取的AI处理能力,第一设备是否支持该AI应用对应的模型的输入部分的处理能力。_
在确定第一设备支持AI应用协同时,可以向第一设备发送能力协商确认消息,该能力协商确认消息可以用于提示第一设备能力协商成功,是否开启AI应用协同功能。
在一些实施例中,当AI应用协同能力协商成功时,第二设备上还可以显示该能力协商确认消息的提醒消息,如图4d中的(a)所示,以提示用户是否开启第一设备与第二设备的AI应用协同功能。在该提醒消息的界面上,可以设置有设置控件,用于跳转至AI应用协同的界面,以使用户可以与第一设备创建AI应用协同。
在一些实施例中,当AI应用协同能力协商失败(例如,第一设备与第二设备的NPU架构不兼容,无法支持同一个AI模型)时,第二设备可以向第一设备返回能力协商失败的消息,如图4d中的(b)和如图4d中的(c)所示,在该提醒消息的界面上,可以设置有查看详情控件,用于跳转至AI应用协同的界面,以使用户可以查看具体的AI应用协同能力协商失败结果的界面。此时,用于可以根据协商失败结果,确定是否仍与第一设备传输媒体信息,以实现第二设备的AI应用功能。另一种可能的方式,如图4d中的(b)所示,在该提醒消息的界面上,还可以提示用户是否与第一设备传输媒体信息,以实现第二设备的AI应用功能。
一种的可能的场景,第一设备和第二设备可以协商传输媒体信息的能力。通过相应的握手协议,确定第一设备、第二设备及相应的传输接口同时支持传输的媒体信息的传输方式。从而,第一设备可以与第二设备建立传输媒体信息的通信链路,第一设备可以基于获取的媒体信息,根据第一设备和第二设备支持的媒体编码方式,第一设备可以将媒体信息进行媒体编码和信道编码,将编码后的媒体信息的信号发送给第二设备。此时,第二设备根据接收到的媒体信息的信号,进行解码,获得相应的媒体信息,并根据AI应用的需要,对媒体信息进行处理。当媒体信息需要在第二设备进行显示/播放时,第二设备可以对该媒体信息进行显示/播放。当媒体信息的处理结果需要在第一设备进行显示/播放时,第二设备可以将媒体信息的处理结果进行媒体编码和信道编码,生成可以使得第一设备进行显示/播放的媒体信息的信号,从而,第一设备可以接收到该媒体信息的信号,并进行显示/播放。
步骤404:第二设备向第一设备发送认证请求消息。
其中,所述认证请求消息用于请求所述第一设备是否与所述数据处理装置建立可信的通信连接,所述通信连接用于确认所述数据处理装置控制所述第一设备的权限。
在一些实施例中,响应于用户对第一界面的AI应用协同请求的操作,第二设备可以向第一设备发送安全鉴权请求消息;该安全鉴权请求消息用于请求第二设备获取第一设备的媒体采集单元等用于AI应用协同功能的单元的控制权限。
其中,安全鉴权请求对应的鉴权方式可以包括手动授权、统一生物特征认证授权、账户授权、云端授权、近场通信授权等方式。以用户输入用户名和密码的方式举例说明。此时,第二设备可以显示第一设备的认证界面,该认证界面用于提示用户输入登录认证第一设备的用户名和密码。在第二设备接收到用户的确认输入后,第二设备可以将用户输入的用户名和密码携带在安全鉴权请求消息发送至第一设备。第一设备从传输端口接收到第二设备的安全鉴权请求消息,对安全鉴权请求中携带的用户名和密码进行合法性验证。在第一设备对第二设备的合法性验证通过后,第一设备可以向第二设备发送安全鉴权响应消息,该安全鉴权响应消息用于通知第二设备,第二设备是否可以获得第一设备的媒体采集单元等用于AI应用协同功能的单元的控制权限。
步骤405:第一设备向第二设备发送认证响应消息。
其中,所述认证响应消息用于确认所述第一设备是否与所述数据处理装置建立可信的 通信连接。
步骤406:第二设备可以向第一设备发送认证成功消息。
需要说明的是,该认证成功消息是可选的。
其中,所述认证成功消息包括:所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识。
在一些实施例中,响应于第一设备的安全鉴权响应消息,第二设备可以为第一设备与第二设备组成的AI应用协同的分布式系统分配一个分布式系统的标识,即可以认为第一设备预第二设备组成一个超级设备或分布式系统。在该分布式系统中,还可以为第一设备分配相应的设备标识,为第二设备分配相应的设备标识,用于分布式系统之间的数据传输。该分布式系统的标识可以用于第二设备与第一设备建立通信链路。并根据需求建立相应的媒体数据链路、AI应用对应的模型获得的特征数据传输链路,及控制链路。从而,第二设备可以通过特征数据传输链路接收第一设备发送的特征数据,在对特征数据进行处理后,可用于AI应用。第二设备可以通过媒体数据链路接收第一设备发送的媒体信息。第二设备还可以通过控制链路,向第一设备发送控制指令,该控制指令可以用于指示对第一设备中的单元(媒体信息采集单元、存储单元、媒体信息播放单元等)的控制,例如对媒体信息采集单元采集媒体信息的开始和结束,对媒体信息采集单元的参数调节控制和操作等控制方式。通过第一设备和第二设备的协同,对第一设备采集的媒体信息实现相应的AI应用的任务。
当安全鉴权失败时,第一设备无法获得第一设备相应AI应用协同所需单元的控制权限,可以向第二设备显示第二设备与第一设备建立AI应用协同失败的通知消息。
当安全鉴权部分失败时,例如,第一设备中的媒体采集单元鉴权成功,存储单元鉴权失败,此时,可以针对鉴权成功的单元与第二设备建立AI应用协同的通信连接。进一步的,还可以在第二设备的AI应用协同的界面,显示第二设备与第一设备的媒体采集单元鉴权成功,建立AI应用协同成功的通知消息。再比如,第一设备的AI处理单元鉴权成功,此时,第二设备可以与第一设备的AI处理单元建立AI应用协同的通信连接。即第二设备可以为第二设备与第一设备的AI处理单元配置设备标识,组成相应的AI应用协同的分布式系统。该设备用于第一设备与第二设备建立特征数据的通信链路,还可以建立媒体信息的通信链路。第一设备可以向第二设备发送特征数据,使得第二设备根据第一设备发送的特征数据,对AI应用的模型的输出部分进行运算,得到AI应用的推理结果,从而实现第一设备与第二设备的AI应用协同。
在一些实施例中,第一设备和第二设备都可以为多个,可以由一个第一设备发起AI应用协同的通信连接的建立,也可以是多个设备共同向服务器发起能力协商请求和安全鉴权请求,此时,可以通过服务器对第一设备和第二设备分别进行上述建立AI应用协同的能力的确认,及对第一设备和第二设备进行安全鉴权,在能力协商成功,及鉴权成功后,可以为第二设备配置相应的第一设备的控制权限,并为组成的分布式系统配置设备标识。
当AI应用协同组成的分布式系统中存在多个第一设备或多个用于AI应用协同的第一设备的媒体信息采集单元、存储单元、传输接口等多个单元时,可为不同的单元分配该设备标识的单元标识。如图4b中的(c)所示,为第二设备与第一设备成功建立AI应用协同后的显示界面,在该显示界面中,可以显示可以与第二设备建立AI应用协同的单元, 及单元标识,单元的状态(是否开启AI应用协同)。在另一些实施例中,还可以通过用户主动设置与第二设备建立AI应用协同的单元,不申请不做限定。
从而,多个第一设备或多个媒体信息采集单元采集到的媒体信息,经各自的模型的输入部分进行特征提取后,获得的特征数据,可以根据各自的单元标识,以区分各自的特征数据的来源,从而,通过传输接口聚合后统一发送给第二设备,第二设备可以根据各自的单元标识,确定各自的特征数据,从而,进行相应的处理,并通过携带各自的单元标识向对应的设备发送控制指令,以实现多设备的AI应用协同,便于第二设备对多个第一设备或多个单元进行统一的控制、管理。
在本申请其他一些实施例中,在第一设备对第二设备的合法性验证通过后,第二设备还可以显示通知消息,以通知用户第二设备与第一设备已成功建立连接,其可以建立与第一设备中的媒体数据用于与第二设备AI应用协同的传输场景。
进一步的,在第一设备与第二设备建立通信连接的过程中,为避免用户对AI应用协同功能的频繁操作,一种可能的实现方式,第二设备可以预先设置开启AI应用协同功能的触发条件,例如,在第一设备与第二设备建立通信连接,且第二设备开启了相应的AI应用时,第二设备自动开启AI应用协同功能。此处的触发AI应用协同功能的AI应用,可以通过设置白名单的方式确定,如图4b所示,白名单可以是用户在AI应用协同功能的界面设置的,也可以通过出厂设置默认设置,在此不做限定。
相应的,第一设备也可以预先设置开启AI应用协同功能的触发条件,例如,在第一设备与第二设备建立通信连接后,第一设备自动开启AI应用协同功能。
另一种可能的实现方式,考虑到该通信连接是用于实现AI应用协同功能的,因此,用户可以在第二设备中的AI应用协同控制界面设置:查找的第一设备为开启了AI应用协同功能的第一设备。从而,使得第二设备仅与开启了AI应用协同功能的第一设备建立通信连接,避免建立不必要的通信连接,浪费网络资源。示例性的,在第二设备确定第一设备开启AI应用协同功能时,与第一设备建立通信连接。
在第一设备与第二设备建立AI应用协同功能的通信连接后,可以通过建立的通信连接,确认第一设备与第二设备相应加载AI算法模型的输入部分和输出部分。其中,第一设备具备媒体信息获取能力,并能够加载AI算法模型的输入部分;第二设备能够加载AI算法模型的输出部分;每个第一设备和第二设备可以加载一个或多个AI算法模型的输入部分和输出部分。
为保证第一设备和第二设备加载的AI算法模型的输入部分和输出部分可以用于当前AI应用协同功能。一种可能的实现方式,可以通过第一设备与第二设备之间的能力协商,根据AI应用对应的AI算法模型,确定第一设备和第二设备是否各自加载了相应的AI算法模型的输入部分和输出部分。
在一些实施例中,第二设备还可以通过该能力协商请求消息,向第二设备询问第一设备的所加载的AI算法模型的模型ID和版本ID。相应的,第一设备在接收到第二设备发送的能力协商请求后,可以向第二设备返回能力协商响应,该能力协商响应消息中可以包括:第一设备所加载的AI算法模型的模型ID和版本ID。
在一些实施例中,该AI算法模型的模型ID和版本ID,可以在AI应用协同界面上显示,并显示加载状态。例如,如图4e中的(a)所示,第一设备的单元1用于与第二设备建立针对第一应用的AI应用协同,此时,可以在单元1的显示栏中显示第一设备加载的第一 应用对应的模型ID和版本ID。
其中,以算法模型存储在第二设备上为例。可以通过判断第一设备加载的AI算法模型的输入部分对应的模型ID、版本ID,与第二设备加载的AI算法模型的输出部分对应的模型ID、版本ID是否一致或可兼容来确定。
当第二设备确定第一设备的加载的AI算法模型的输入部分对应的模型ID、版本ID,与第二设备加载的AI算法模型的输出部分对应的模型ID、版本ID一致或可兼容时,可以确认第一设备和第二设备的AI应用协同的建立完成。
当第二设备确定第一设备的加载的AI算法模型的输入部分对应的模型ID、版本ID,与第二设备加载的AI算法模型的输出部分对应的模型ID、版本ID不一致或不兼容时,第二设备可以确定第一设备的版本需要更新,此时,第二设备可以在显示屏上显示加载失败的界面,进一步的,还可以在显示界面显示提示框,该提示框用于提示用户是否更新第一设备的模型或模型的版本。响应于更新AI算法模型的输入部分的指令,第二设备可以在显示屏上显示更新界面,例如,如图4e中的(b)所示,第一设备的单元1用于与第二设备建立针对第一应用的AI应用协同,此时,可以在单元1的显示栏中显示第一设备加载的第一应用对应的模型ID和版本ID需要更新。此时,可以向第一设备发送与第二设备加载的AI算法模型的输出部分对应的模型ID、版本ID一致或可兼容的AI算法模型的输入部分,并在第一设备上加载该AI算法模型的输入部分。可选的,如图4e中的(b)所示,还可以在显示屏上显示的更新界面中显示更新的进度或状态,在确认第一设备上加载该AI算法模型的输入部分完成时,可以确认第一设备和第二设备的AI应用协同的建立。
上述以第二设备存储有AI算法模型的输入部分和输出部分为例进行了说明,相应的,第一设备存储有AI算法模型的输入部分和输出部分时,也可以参考上述实施例,以完成第一设备与第二设备相应加载AI算法模型的输入部分和输出部分。另外,AI算法模型的输入部分和输出部分还可以存储在服务器中,此时,可以通过服务器完成第一设备与第二设备相应加载AI算法模型的输入部分和输出部分的过程,在此不再赘述。
对于一个AI应用,其输入部分和输出部分可以一一对应的,也可以是多个输入部分对应一个输出部分,即一个AI算法模型,包括多个输入部分和一个输出部分。一个输入部分对应一个第一设备,一个输出部分对应一个第二设备。
例如,利用多个摄像头获取的视频信息进行机器视觉处理的场景,此时,多个摄像头可以作为多个第一设备,将获得机器视觉处理结果的设备作为第二设备。再比如,利用摄像头和激光雷达等多模式的传感器获取的数据信息用于环境理解等应用;此时,摄像头可以作为1个第一设备,激光雷达作为1个第一设备,每个传感器作为一个第一设备,将获得环境理解等应用的处理结果的设备作为第二设备。此时,一个或多个第一设备上的多个媒体信号可以经过多个不同的神经网络处理单元上加载的多个AI算法模型输入部分进行处理。即每个第一设备将获取到的媒体信息进行特征提取,将特征提取后的特征数据发送给第二设备,使得第二设备可以通过获取到的多个第一设备的传感器的媒体信息对应的特征数据。
一种可能的实现方式,一个或多个第一设备上获得的多个特征数据可以进行独立的打包封装,在传输接口中进行聚合后统一传输,此时,第二设备可以根据接收到的数据包,进行恢复,以确定出各个第一设备对应的特征数据,将各个第一设备的特征数据同步后输入到第二设备的神经网络处理单元上加载的AI算法模型输出部分进行处理,以获得推理 结果,用于后续的AI应用,从而,实现多设备间的协同AI任务。
另一种可能的实现方式,一个或多个第一设备可以分别与第二设备建立通信链路,从而,使得各个第一设备可以分别向第二设备发送相应的特征数据。此时,第二设备可以根据接收到的各第一设备发送的特征数据输入到相应的第二设备对应的AI算法模型的输出部分进行处理,以获得AI算法模型的推理结果,实现多设备间的协同AI任务。
示例一
在该示例一中,可以用于分体电视、分体式AR/VR等场景。第一设备可以为分体电视的屏幕端、AR/VR头戴式显示设备等设备。第二设备可以为分体电视主机盒子、手机、PC、游戏主机等。如图5a所示,为本示例对应的系统架构的示意图。第一设备和第二设备通过相应的传输接口,建立AI应用协同的通信连接,以组成AI应用协同的分布式系统。在该场景下,如图5b所示,本申请实施例的媒体信息传输方法的流程可以包括以下步骤:
步骤501:第一设备获取媒体信息。
其中,第一设备可以是具备音视频的采集和显示播放能力以及AI算法模型处理硬件。
通过第一设备的音视频采集单元采集原始音视频信号,经过第一设备的处理单元的预处理后,获得待传输的媒体信息,并传输至第一设备的NPU。例如,如图5c所示,第一设备为分体电视的屏幕端,在第一设备的摄像头上采集的人的多个视频图像帧作为待传输的媒体信息。
步骤502:第一设备根据AI算法模型的输入部分(例如,第一特征提取模型),对媒体信息进行特征提取,确定特征数据。
第一设备根据当前确定协同处理的AI应用(例如,应用1)确定第一设备的NPU中加载有第一应用的AI算法模型的输入部分,从而,通过第一设备中的NPU中的第一应用的AI算法模型的输入部分,对媒体信息进行特征提取,获得特征数据。以第一应用为机器视觉交互的相关应用为例,例如,如图5d所示,第一应用对应的AI算法模型可以是用于对人的手势进行识别的AI算法模型,此时,可以根据手势识别的AI模型的输入部分,对在第一设备的摄像头上采集的人的多个视频图像帧进行特征提取,从而,确定进行特征提取后的特征数据。
步骤503:第一设备向第二设备发送特征数据。
将特征数据传输到第二设备的NPU中。第二设备具备媒体信息处理和控制能力,例如,具备AI算法模型处理硬件以及媒体信息处理和人机交互能力。
步骤504:第二设备根据AI算法模型的输出部分(例如,第一特征数据处理模型),对特征数据进行处理。
利用AI算法模型的输出部分进行处理,第二设备可以得到AI算法模型的推理结果。AI算法模型的推理结果提供给后续的AI应用程序,获得后续语音交互、机器视觉交互、环境建模等AI应用的任务的处理结果。
在一些实施例中,第二设备获得的AI应用的处理结果需要在第一设备的显示界面上显示,例如,如图5d所示,在第一设备上采集的人的多个视频图像帧作为待传输的媒体信息,此时,可以根据手势识别的AI模型的输出部分,对获取的特征数据进行处理,以识别出第一设备采集的人的视频图像中的手势。识别出的手势可以通过在图像的相应位置生成识别框作为该AI应用的处理结果。此时,可以将该AI应用的处理结果显示到第一设 备的显示屏上,因此,第二设备可以向第一设备发送该手势识别结果的控制指令,以指示第一设备在显示屏的相应位置上显示该手势识别结果,从而,使得用户可以根据显示的手势识别结果确定手势识别成功。
可选的,步骤505a:第二设备向第一设备发送第一消息。
其中,所述第一消息用于指示所述第一设备采集媒体数据的状态,所述第一设备采集媒体数据的状态包括以下至少一项:开启状态、关闭状态或采集媒体数据的参数。
步骤506a:响应于所述第一消息,调整所述第一设备采集媒体数据的状态。
在另一些实施例中,当第二设备拥有第一设备的摄像头的控制权限时,第二设备还可以根据手势识别结果,确定第一设备的摄像头的参数设置是否合理,当确定第一设备的摄像头的参数设置需要调整时,可以生成相应的控制指令,该控制指令用于调整第一设备的摄像头的参数(例如,第一设备的摄像头的位姿,摄像头的焦距,焦点位置等,开启的摄像头的类型等),在使得用户使用机器视觉应用过程中,保证第一设备的摄像头可以随着用户的手势、姿态或位置的变化相应调整摄像头的参数,从而实现对手势的实时跟踪识别。
可选的,步骤505b:第二设备向第一设备发送第二消息。
在一些实施例中,第二消息用于指示第一设备获取第一数据。
步骤506b:第一设备获取第一数据。
第一设备响应于所述第二消息,第一设备可以基于网络或者第一设备的存储单元或者获取所述第一数据,或者,第一设备的媒体信息采集单元采集所述第一数据。从而,第一设备可以向所述第二设备发送所述第一数据;所述第一数据可以为:所述第一设备采集到的媒体数据,所述第一设备存储的数据,第一设备接收的数据。
再比如,当第二设备需要控制第一设备的媒体信息采集单元,以调整媒体信息采集单元的参数时,可以先向第一设备发送第二消息,该第二消息可以用于请求获取第一设备的媒体信息采集单元的参数。
在另一些实施例中,第二设备可以基于AI应用的推理结果,控制第一设备采集媒体信息,此时,所述第二消息可以用于指示第一设备采集第三媒体数据的特征数据。例如,第二设备根据AI应用的推理结果,确定还需要通过第一设备采集音频信息,此时,可以通过第二消息指示第一设备采集相应的音频信息,可选的,第二消息还可以指示采集该音频信息后对该音频信息进行特征提取的模型ID和版本ID。此时,第一设备响应于所述第二消息,采集所述第三媒体数据,并根据模型ID和版本ID对应的特征提取模型对所述第三媒体数据进行特征提取,得到第三特征数据,向第二设备发送所述第三特征数据。从而,第二设备在接收到第三特征数据后,根据模型ID和版本ID对应的特征数据处理模型对第三特征数据进行处理,以获得音频信息对应的AI应用的推理结果。从而,更好的实现AI应用的任务。
在一些实施例中,所述第二设备还可以根据接收到的用户的操作信息和/或AI模型的处理结果,生成向第一设备发送的第三消息。
举例来说,用户可以对AI应用(例如,应用1)的界面进行操作,该操作可以为点击操作,也可以为手势操作,也可以为语音指令操作,在此不做限定。第二设备可以接收来自第一设备采集的用户对AI应用的界面的操作指令,从而,根据该操作指令,及AI模型的处理结果,生成对第一设备的第三消息。
可选的,步骤505c:第二设备向第一设备发送第三消息。
其中,所述第三消息为所述第二设备根据所述第一特征数据确定的;所述第三消息用于指示所述第一设备显示的内容。
步骤506c:第一设备响应于所述第三消息,通过显示单元显示所述第三消息中用于指示所述第一设备显示的内容。
例如,以用户的手势操作指令为开启相应的AI应用时,第二设备通过识别该手势,确定该手势操作指令为开启相应的AI应用,此时,第二设备根据该操作指令,开启该AI应用,并向第一设备发送显示该AI应用的开启界面所需的媒体信息。此时,第一设备接收到该媒体信息后,可以通过第一设备的显示屏显示该AI应用的开启界面。
再比如,以用户的操作指令为跳转到对应操作的视频界面并显示相应的视频,此时,第二设备根据识别到的用户的操作指令,获取相应的视频的媒体信息,从而,通过传输接口,将该媒体信息发送给第一设备,并在第一设备的显示屏上显示播放,以实现对用户的操作指令的响应,完成人机交互。
示例二
如图6a所示,为示例二的一种系统架构的示意图。第二设备具备显示屏。此时,第一设备可以为外置摄像头配件、车载摄像头、家用监控摄像头、具备视频采集能力的智能家电、分体电视的屏幕端、AR/VR头显等终端设备。第二设备可以为分体电视主机盒子、手机、车载主机、PC、游戏主机等具备较强计算显示的终端设备。第一设备和第二设备通过相应的传输接口,建立AI应用协同的通信连接,以组成AI应用协同的分布式系统。在该场景下,如图6b所示,本申请实施例的媒体信息传输方法的流程可以包括以下步骤:
步骤601:第一设备获取媒体信息。
其中,第一设备可以是具备音视频的采集能力以及AI算法模型处理硬件。
通过第一设备的音视频采集单元采集原始音视频信号,经过第一设备的处理单元的预处理后,获得待传输的媒体信息,并传输至第一设备的NPU。例如,如图6c所示,以第一设备为车载设备中的传感器单元为例,在第一设备的摄像头上采集的车辆行驶过程中的道路和车辆外部的视频图像作为待传输的媒体信息。
步骤602:第一设备根据AI算法模型的输入部分(例如,第一特征提取模型),对媒体信息进行特征提取,确定特征数据。
第一设备的NPU中加载有AI算法模型的输入部分,从而,通过NPU中的AI算法模型的输入部分,对媒体信息进行特征提取,获得特征数据。例如,AI应用为自动驾驶的相关应用,此时,AI应用对应的AI算法模型可以是用于对车辆行驶的车道、路况等环境进行识别的AI算法模型,此时,可以根据道路识别的AI模型的输入部分,对在第一设备的传感器(例如,雷达传感器、摄像头等)上采集的车道的图像进行特征提取,从而,确定进行特征提取后的特征数据。例如,如图6c所示,第一设备上用于AI应用协同的媒体信息采集单元可以为传感器单元1(例如,雷达传感器)。此时,可以通过传感器单元1采集到的媒体信息,通过传感器单元1相应的第一特征提取模型进行特征提取,得到特征数据1。
步骤603:第一设备向第二设备发送特征数据。
将特征数据传输到第二设备的NPU中。第二设备具备媒体信息处理和控制能力,例如,具备AI算法模型处理硬件以及媒体信息处理和人机交互能力、显示能力等。
步骤604:第二设备根据AI算法模型的输出部分(例如,第一特征数据处理模型),对特征数据进行处理。
利用AI算法模型的输出部分进行处理,第二设备可以得到AI算法模型的推理结果。AI算法模型的推理结果提供给后续的AI应用程序,获得后续语音交互、机器视觉交互、环境建模等AI应用的任务的处理结果。
结合上述例子,在步骤604中,根据第一设备的雷达传感器采集的车道的点云图像进行特征提取后的特征数据1,输入至相应的识别车道的AI算法模型的输出部分进行处理,可以得到图像中的车道信息(如图6c所示,可以确定车辆即将驶入从左到右的第2车道),从而,可以为第二设备提供更准确的第一设备的定位信息(即车辆所在的车道),从而可以根据第一设备的定位信息,为第一设备提供更好的导航路径等应用。
在另一些实施例中,第一设备可以是多种类型的媒体信息采集设备,用于对AI应用提供更多的媒体信息,从而获得更好的AI算法的结果。仍以车辆中的多种类型的传感器为例,例如,在不同天气状况下,仅通过单一的传感器对当前的道路进行识别可能出现较大的误差,此时,可以通过多种类型的传感器采集到的媒体信息进行综合识别,从而可以获得更好的道路识别效果。
例如,采集的媒体信息可以是一个视频/图像信号也可以是多个视频/图像信号的组合;其中每个视频/图像信号可以是可见光图像,也可以是红外图像、雷达信号、深度信息等其他模态的视频/图像信号;此时,一个或多个第一设备上的多个媒体信息采集单元可以经过多个不同的NPU加载的多个AI算法模型的输入部分进行处理,提取出相应的特征数据。例如,如图6c所示,第一设备上用于AI应用协同的媒体信息采集单元可以包括传感器单元1(例如,雷达传感器)和传感器单元2(例如,摄像头)。此时,可以通过传感器单元1采集到的媒体信息,通过传感器单元1相应的第一特征提取模型进行特征提取,得到特征数据1。通过传感器单元2采集传感器到的媒体信息,通过传感器单元2相应的第一特征提取模型进行特征提取,得到特征数据2。从而,还可以将各NPU输出的多个特征数据进行独立的打包封装,在传输接口中进行聚合后统一传输到第二设备。第二设备根据接收到的特征数据1和特征数据2,将各个特征数据1和特征数据2输入到各自的AI算法模型的输出部分进行处理,或者,输入到对特征数据1和特征数据2进行融合处理的AI算法模型,以获得更好的识别效果。
步骤605:在第二设备上显示AI应用的处理结果。
在一些实施例中,AI应用的处理结果,可以在第二设备上显示。例如,如图6d所示,可以在第二设备的显示屏上显示当前车辆所在的车道,显示基于车辆所在车道为用户规划的导航路径等。第一设备上用于AI应用协同的媒体信息采集单元可以包括传感器单元1(图中标记为1)和传感器单元2(图中标记为2).
根据AI应用的任务的处理结果,也可以生成对第一设备的控制指令,具体方式可以参考图5b中的实施方式,在此不再赘述。
第一设备和第二设备通过传输接口连接组成AI应用协同的分布式系统,将第一设备的信息感知能力、轻量化的AI处理能力与第二设备更强大的计算硬件、AI处理能力和交互能力相结合,协同完成语音交互、视觉交互、环境建模等AI应用的任务。
示例三
如图7a所示,第一设备和第二设备可以组成AI应用协同的分布式语音交互系统。其中,第一设备可以具备音频信息采集能力以及AI算法模型处理硬件。例如,智能耳机、智能音箱、AR/VR头显、车载音频采集设备、具备音频采集能力的智能家电。第二设备可以为手机、智能电视、车载主机等具备较强计算的终端设备。可选的,第二设备还可以具有显示功能。考虑到该示例中,主要用于传输音频信号,因此,该AI应用协同对应的传输接口,可以是wifi、蓝牙等无线传输系统,也可以是有线传输的电信号、光纤传输的光信号等。如图7b所示,本申请提供一种媒体信息传输方法的流程示意图,具体包括:
步骤701:第一设备获取媒体信息。
其中,第一设备可以是具备音频的采集能力以及AI算法模型处理硬件。
通过第一设备的音频单元采集原始音频信号,经过第一设备的处理单元的预处理后,获得待传输的媒体信息,并传输至第一设备的NPU。例如,如图7c所示,第一设备为智能耳机,通过第一设备中的麦克风单元与第二设备建立AI应用协同(例如,降噪、语音交互等AI应用),在第一设备的麦克风上采集的用户输入的语音或者环境噪音等媒体信息。
步骤702:第一设备根据AI算法模型的输入部分(例如,第一特征提取模型),对媒体信息进行特征提取,确定特征数据。
第一设备的NPU中加载有AI算法模型的输入部分,从而,通过NPU中的AI算法模型的输入部分,对媒体信息进行特征提取,获得特征数据。例如,AI应用为语音识别交互的相关应用,此时,AI应用对应的AI算法模型可以是用于对语音进行识别的AI算法模型,此时,可以根据语音识别的AI模型的输入部分,对在第一设备的麦克风上采集的音频信息进行特征提取,从而,确定进行特征提取后的特征数据。再比如,AI应用为自动降噪的相关应用,此时,AI应用对应的AI算法模型可以是用于对环境噪音进行识别的AI算法模型,此时,可以根据噪音识别的AI模型的输入部分,对在第一设备的麦克风上采集的音频信息进行特征提取,从而,确定进行特征提取后的特征数据。
步骤703:第一设备向第二设备发送特征数据。
将特征数据传输到第二设备的NPU中。第二设备具备媒体信息处理和控制能力,例如,具备AI算法模型处理硬件以及媒体信息处理和人机交互能力、显示能力等。
步骤704:第二设备根据AI算法模型的输出部分(例如,第一特征数据处理模型),对特征数据进行处理,确定第一应用的结果。
可选的,步骤705:第二设备向第一设备发送第一应用的结果。
步骤706:第一设备播放第一应用的结果。
利用AI算法模型的输出部分进行处理,第二设备可以得到AI算法模型的推理结果。AI算法模型的推理结果提供给后续的AI应用程序,获得后续语音识别、自然语言处理、声纹识别等语音交互任务的处理结果。
结合上述例子,可以根据对特征数据的处理,确定第一设备采集的语音识别结果,例如,该语音识别结果为搜索指定的视频。此时,如图7d所示,第二设备可以在第二设备的显示界面显示该语音识别结果,并根据语音识别结果,搜索相应的视频,进一步的,还可以在AI协同界面上或者,跳转到相应的视频播放应用中显示搜索到的视频,以完成语音识别交互的AI应用的任务。
结合上述例子,可以根据对特征数据的处理,确定第一设备采集的噪音识别结果,从而根据噪音识别结果,生成相应的降噪的音频信息,将该降噪的音频信息通过传输接口发 送给第一设备,使得第一设备的麦克风在进行录音或者音视频播放单元进行音频播放时,通过该降噪的音频信息实现对第一设备的录音或音频播放的降噪。
第一设备获取的音频信息可以不传输到第二设备,而是进过AI算法模型的输入部分转换为抽象的特征数据后再传输;特征数据经过模型处理后存在明显的信息损失无法恢复为人可以直接理解的音视频信息,提升隐私保护能力;特征数据的数据量明显低于原始的音视频信息,在较小的信道带宽条件下也可以实时传输,省去了额外的压缩编码过程,降低系统功耗和延时,降低成本提升产品竞争力。
将发送端的音频感知能力、轻量化的AI处理能力与第二设备更强大的计算硬件、AI处理能力相结合,协同完成语音识别、自然语音处理、声纹识别模等语音交互任务;接口不需要支持视频信息的双向传输,所需带宽低,更适合短距无线传输。
本申请的上述各实施方式可以任意进行组合,以实现不同的技术效果。
上述本申请提供的实施例中,从第一设备和第二设备作为执行主体的角度对本申请实施例提供的方法进行了介绍。为了实现上述本申请实施例提供的方法中的各功能,电子设备可以包括硬件结构和/或软件模块,以硬件结构、软件模块、或硬件结构加软件模块的形式来实现上述各功能。上述各功能中的某个功能以硬件结构、软件模块、还是硬件结构加软件模块的方式来执行,取决于技术方案的特定应用和设计约束条件。
基于相同的构思,图8所示为本申请的一种电子设备800,包括:收发模块801、采集模块803和处理模块802,可选的,电子设备800还可以包括显示模块。示例性的,电子设备800可以为本申请实施例中的第一设备。此时,收发模块801包括第一传输接口。
采集模块803,用于采集第一媒体信息;
处理模块802,用于对所述第一媒体信息进行特征提取,确定所述第一媒体信息的第一特征数据;通过所述第一传输接口向第二设备发送所述第一特征数据,所述第一特征数据用于所述第二设备获得第一应用的结果。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收所述第二设备发送的能力协商请求消息;所述能力协商请求消息用于请求所述第一设备支持的传输协议,及所述第一设备的特征提取能力;所述第一设备的传输协议用于指示所述第一设备支持传输特征数据;所述第一设备的特征提取能力用于指示所述第一设备支持提取所述第一媒体信息的第一特征数据;通过所述第一传输接口向所述第二设备发送能力协商响应消息;所述能力协商响应消息用于确认所述第一设备支持传输特征数据的传输协议及所述第一设备的特征提取能力。
一种可能的实现方式,处理模块802,用于响应于在第一应用上的第一操作,通过收发模块801向所述第二设备发送第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;收发模块801,用于接收所述第二设备返回的第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启第一应用协同。
一种可能的实现方式,收发模块801,用于接收来自所述第二设备发送第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;处理模块802,用于响应于在第一应用上的第三操作,通过收发模块801向所述第二设备发送第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启第一应用协同。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口向所述第二设备发送的能力协商请求消息;所述能力协商请求消息用于请求所述第二设备支持的传输协议,及所述第二设备的特征数据处理能力,所述第二设备的传输协议用于指示所述第二设备支持传输特征数据;所述第二设备的特征数据处理能力用于指示所述第二设备支持处理所述第一特征数据获得第一应用的结果的能力;通过所述第一传输接口接收来自所述第二设备的能力协商响应消息;所述能力协商响应消息用于确认所述第二设备支持传输特征数据的传输协议及所述第二设备的特征数据处理能力。
一种可能的实现方式,处理模块802,用于通过收发模块801获取第一特征提取模型;其中,所述第一特征提取模型用于对所述第一媒体信息进行特征提取,所述第一特征提取模型的版本与第一特征数据处理模型的版本对应,所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果。
一种可能的实现方式,所述能力协商响应消息还包括:所述第一设备中的特征提取模型的版本;或者,所述第二设备中的特征数据处理模型的版本。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收来自所述第二设备的第一特征提取模型,或者,接收来自服务器的第一特征提取模型,或者,读取所述第一设备存储的第一特征提取模型。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口向所述第二设备发送第一特征数据处理模型;其中,所述第一特征提取模型的版本与所述第一特征数据处理模型的版本对应,所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果。
一种可能的实现方式,处理模块802,用于通过收发模块801获取第二特征提取模型,所述第二特征提取模型的版本与第二特征数据处理模型的版本对应,所述第二特征提取模型和所述第二特征数据处理模型为更新所述第一特征提取模型和所述第二特征数据处理模型后确定的。
一种可能的实现方式,处理模块802,用于根据所述第一特征提取模型,对训练样本进行特征提取,生成第一训练特征数据;收发模块801,用于通过所述第一传输接口向所述第二设备发送所述第一训练特征数据;所述第一训练特征数据用于训练所述第一特征提取模型和所述第一特征数据处理模型。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收来自所述第二设备的反馈数据,所述反馈数据为所述第二设备根据所述第一训练特征数据训练后确定的;所述反馈数据用于所述第一设备训练所述第一特征提取模型。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收来自所述第二设备的第一消息;所述第一消息用于指示所述第一设备采集媒体信息的状态;响应于所述第一消息,调整所述第一设备采集媒体信息的状态。
一种可能的实现方式,所述第一设备采集媒体信息的状态包括以下至少一项:开启状态、关闭状态或采集媒体信息的参数。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收来自所述第二设备的第二消息;其中,所述第二消息用于指示所述第一设备获取第一数据;处理模块802,用于响应于所述第二消息,获取所述第一数据,或者,采集所述第一数据;向所述第二设备发送所述第一数据;所述第一数据为以下一项:所述第一设备采集到的媒体信息,所述 第一设备的参数,所述第一设备存储的数据,第一设备接收的数据。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口向所述第二设备发送所述第一数据。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收来自所述第二设备的第二消息,所述第二消息用于指示所述第一设备采集第三媒体信息的特征数据;响应于所述第二消息,采集所述第三媒体信息;对所述第三媒体信息进行特征提取,得到第三特征数据;通过所述第一传输接口向所述第二设备发送所述第三特征数据。
一种可能的实现方式,所述第二消息或所述第一消息为所述第二设备根据所述第一特征数据确定的。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收来自所述第二设备的第三消息,所述第三消息为所述第二设备根据所述第一特征数据确定的,所述第三消息用于指示所述第一设备显示的内容;处理模块802,用于响应于所述第三消息,通过显示模块显示所述第三消息中用于指示所述第一设备显示的内容。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收所述第二设备发送的认证请求消息,所述认证请求消息用于请求所述第一设备是否与所述第二设备建立通信连接,所述通信连接用于确认所述第二设备控制所述第一设备的权限;通过所述第一传输接口向所述第二设备发送认证响应消息;所述认证响应消息用于确认所述第二设备控制所述第一设备的权限。
一种可能的实现方式,收发模块801,用于通过所述第一传输接口接收所述第二设备发送的认证成功消息;所述认证成功消息包括:所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识。
一种可能的实现方式,电子设备800还可以包括第一模块;所述认证成功消息还包括以下至少一项:所述第一设备的第一模块的标识,及所述第一模块在所述分布式系统中的标识。
一种可能的实现方式,收发模块801还可以包括第三传输接口;所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第一设备发送的特征数据或消息为通过所述第一传输接口封装为第一比特流数据后,通过所述第三传输接口发送的。
一种可能的实现方式,所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第一设备接收的消息为通过所述第三传输接口接收的第二比特流数据,并通过所述第一传输接口将所述第二比特流数据解封装后得到的。
基于相同的构思,图9所示为本申请的一种电子设备900,包括:收发模块901和处理模块902,可选的,电子设备900还可以包括显示模块。示例性的,电子设备900可以为本申请实施例中的第二设备。此时,收发模块901包括第二传输接口。
收发模块901,用于通过所述第二传输接口接收来自第一设备的第一特征数据;所述第一特征数据为根据第一设备对采集的第一媒体信息进行特征提取后确定的;
处理模块902,用于对所述第一特征数据进行处理,得到第一应用的处理结果。
一种可能的实现方式,处理模块902,用于响应于在第一应用上的第二操作,通过收发模块901向所述第一设备发送第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;接收所述第二设备返回的第一响应消息;所述第一响应消息用于确认所述第一设备 与所述第二设备开启第一应用协同。
一种可能的实现方式,处理模块902,用于通过收发模块901接收第一设备发送的第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;响应于在第一应用上的第四操作,通过收发模块901向所述第一设备发送第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启第一应用协同。
一种可能的实现方式,通过所述第二传输接口向所述第一设备发送能力协商请求消息;所述能力协商请求消息用于请求所述第一设备支持的传输协议,及所述第一设备的特征提取能力;所述第一设备的传输协议用于指示所述第一设备支持传输特征数据;所述第一设备的特征提取能力用于指示所述第一设备支持提取所述第一媒体信息的第一特征数据;通过所述第二传输接口接收所述第一设备发送的能力协商响应消息;所述能力协商响应消息用于确认所述第一设备支持传输特征数据的传输协议。
一种可能的实现方式,所述接收来自第一设备的第一特征数据之前,还包括:
收发模块901,用于通过所述第二传输接口接收所述第一设备发送的能力协商请求消息;所述能力协商请求消息用于请求所述第二传输接口支持的传输协议,及所述第二设备的特征数据处理能力,所述第二设备的传输协议用于指示所述第二设备支持传输特征数据;所述第二设备的特征数据处理能力用于指示所述第二设备支持处理所述第一特征数据获得第一应用的结果的能力;通过所述第二传输接口向所述第一设备发送能力协商响应消息;所述能力协商响应消息用于确认所述第二设备支持传输特征数据的传输协议及所述第二设备的特征数据处理能力。
一种可能的实现方式,处理模块902,用于获取第一特征数据处理模型;所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果;第一特征提取模型的版本与所述第一特征数据处理模型的版本对应,所述第一特征提取模型用于对所述第一媒体信息进行特征提取。
一种可能的实现方式,所述能力协商响应消息还包括:所述第一设备中的特征提取模型的版本;或者,所述第二设备中的特征数据处理模型的版本。
一种可能的实现方式,收发模块901,用于通过所述第二传输接口接收来自所述第一设备的第一特征数据处理模型,或者,接收来自服务器的第一特征数据处理模型,或者,处理模块902,用于读取所述第二设备存储的第一特征数据处理模型;
一种可能的实现方式,收发模块901,用于通过所述第二传输接口向第一设备发送第一特征提取模型;其中,所述第一特征提取模型的版本与所述第一特征数据处理模型的版本对应,所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果。
一种可能的实现方式,处理模块902,用于获取第二特征数据处理模型,所述第二特征数据处理模型的版本与所述第二特征提取模型的版本对应,所述第二特征提取模型和所述第二特征数据处理模型为更新所述第一特征提取模型和所述第二特征数据处理模型后确定的。
一种可能的实现方式,处理模块902,用于通过收发模块901接收第一训练特征数据;所述第一训练特征数据为所述第一设备根据所述第一特征提取模型对训练样本进行特征提取后确定的;根据所述第一训练特征数据,训练所述第一特征数据处理模型。
一种可能的实现方式,处理模块902,用于得到第一特征提取模型的反馈数据;所述反馈数据为所述第二设备根据所述第一训练特征数据训练后确定的;所述反馈数据用于所述第一设备训练所述第一特征提取模型;通过收发模块901向所述第一设备发送所述反馈数据。
一种可能的实现方式,通过收发模块901接收所述第二设备发送的第二特征数据;所述第二特征数据为所述第一设备根据采集的第二媒体信息及所述第二特征提取模型进行特征提取后确定的;处理模块902,用于根据所述第二特征数据处理模型对所述第二特征数据进行处理,得到所述第一应用的结果。
一种可能的实现方式,收发模块901用于通过所述第二传输接口向所述第一设备发送第一消息;所述第一消息用于指示所述第一设备采集媒体信息的状态。
一种可能的实现方式,所述第一设备采集媒体信息的状态包括以下至少一项:开启状态、关闭状态或采集媒体信息的参数。
一种可能的实现方式,收发模块901用于通过所述第二传输接口向所述第一设备发送第二消息;所述第二消息用于指示所述第一设备获取第一数据;所述第一数据为以下一项:所述第一设备采集到的媒体信息,所述第一设备的参数,所述第一设备存储的数据,第一设备接收的数据。
一种可能的实现方式,收发模块901用于通过所述第二传输接口接收来自所述第一设备的所述第一数据。
一种可能的实现方式,收发模块901用于通过所述第二传输接口向第一设备发送第二消息;所述第二消息用于指示所述第一设备采集第三媒体信息的特征数据;通过所述第二传输接口接收来自所述第一设备发送的第三特征数据;所述第三特征数据为所述第一设备对采集的第三媒体信息进行特征提取后确定的。
一种可能的实现方式,所述第一消息或所述第二消息为根据所述第一特征数据的处理结果确定的。
一种可能的实现方式,处理模块902,用于响应于所述第一特征数据的处理结果,生成第三消息;所述第三消息用于指示所述第一设备显示的内容。
一种可能的实现方式,所述第一设备的数量为N个;所述方法还包括:
收发模块901用于通过所述第二传输接口接收第四消息;所述第四消息包括所述N个第一设备的M个第一特征数据;N、M为大于1的正整数;M大于或等于N;
处理模块902,用于根据所述M个第一特征数据对应的特征数据处理模型,对所述M个第一特征数据进行处理,得到所述第一应用的结果。
一种可能的实现方式,收发模块901,用于通过所述第二传输接口向所述第一设备发送认证请求消息,所述认证请求消息用于请求所述第一设备是否与所述第二设备建立通信连接;所述通信连接用于确认所述第二设备控制所述第一设备的权限;通过所述第二传输接口接收所述第二设备发送的认证响应消息;所述认证响应消息用于确认所述第一设备是否与所述第二设备建立通信连接。
一种可能的实现方式,处理模块902,用于响应于所述第二设备发送的认证响应消息,为所述第一设备设置所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识;所述第一设备对应的设备标识及所述分布式系统的标识用于所述第一设备和所述第二设备进行通信;所述第一设备通过所述第一传输接口向所述第二设备发 送认证成功消息;所述认证成功消息包括:所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识。
一种可能的实现方式,所述第二设备包括第二模块;所述认证成功消息还包括以下至少一项:所述第二模块的标识,及所述第二模块在所述分布式系统中的标识。
一种可能的实现方式,收发模块901还包括第三传输接口;所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第二设备发送的消息为通过所述第二传输接口封装为第二比特流数据后,通过所述第三传输接口发送的。
一种可能的实现方式,所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第二设备接收的特征数据或消息为通过所述第三传输接口接收的第一比特流数据,并通过所述第二传输接口将所述第二比特流数据解封装后获得的。
本申请实施例还提供一种媒体信息传输系统,包括如图8所示的电子设备800或如图3a所示的第一设备,还包括如图9所示的电子设备900或如图3b所示的第二设备。
本申请实施例还提供一种计算机存储介质,所述计算机可读存储介质用于存储计算机程序,当所述计算机程序在计算机上运行时,使得所述计算机执行如图2a-图7a中的任意一种可能的实施方式中所述的方法。
本申请实施例还提供一种包含指令的计算机程序产品,所述计算机程序产品用于存储计算机程序,当所述计算机程序在计算机上运行时,使得所述计算机执行图2a-图7a中的任意一种可能的实施方式中所述的方法。
应理解,本申请实施例中提及的处理器可以是CPU,还可以是其他通用处理器、数字信号处理器(digital signal processor,DSP)、专用集成电路(application specific integrated circuit,ASIC)、现成可编程门阵列(field programmable gate array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
还应理解,本申请实施例中提及的存储器可以是易失性存储器或非易失性存储器,或可包括易失性和非易失性存储器两者。其中,非易失性存储器可以是只读存储器(read-only memory,ROM)、可编程只读存储器(programmable ROM,PROM)、可擦除可编程只读存储器(erasable PROM,EPROM)、电可擦除可编程只读存储器(electrically EPROM,EEPROM)或闪存。易失性存储器可以是随机存取存储器(random access memory,RAM),其用作外部高速缓存。通过示例性但不是限制性说明,许多形式的RAM可用,例如静态随机存取存储器(static RAM,SRAM)、动态随机存取存储器(dynamic RAM,DRAM)、同步动态随机存取存储器(synchronous DRAM,SDRAM)、双倍数据速率同步动态随机存取存储器(double data rate SDRAM,DDR SDRAM)、增强型同步动态随机存取存储器(enhanced SDRAM,ESDRAM)、同步连接动态随机存取存储器(synchlink DRAM,SLDRAM)和直接内存总线随机存取存储器(direct rambus RAM,DR RAM)。
需要说明的是,当处理器为通用处理器、DSP、ASIC、FPGA或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件时,存储器(存储模块)集成在处理器中。
应注意,本文描述的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
应理解,在本申请的各种实施例中,上述各过程的序号的大小并不意味着执行顺序的先后,各过程的执行顺序应以其功能和内在逻辑确定,而不应对本申请实施例的实施过程构成任何限定。
本领域普通技术人员可以意识到,结合本文中所公开的实施例描述的各示例的单元及算法步骤,能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行,取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能,但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到,为描述的方便和简洁,上述描述的系统、装置和单元的具体工作过程,可以参考前述方法实施例中的对应过程,在此不再赘述。
在本申请所提供的几个实施例中,应该理解到,所揭露的系统、装置和方法,可以通过其它的方式实现。例如,以上所描述的装置实施例仅仅是示意性的,例如,所述单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,例如多个单元或组件可以结合或者可以集成到另一个系统,或一些特征可以忽略,或不执行。另一点,所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口,装置或单元的间接耦合或通信连接,可以是电性,机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外,在本申请各个实施例中的各功能单元可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括:U盘、移动硬盘、只读存储器(read-only memory,ROM)、随机存取存储器(random access memory,RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述,仅为本申请的具体实施方式,但本申请实施例的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本申请实施例揭露的技术范围内,可轻易想到变化或替换,都应涵盖在本申请实施例的保护范围之内。因此,本申请实施例的保护范围应所述以权利要求的保护范围为准。
Claims (29)
- 一种媒体信息传输方法,其特征在于,应用于第一设备,所述第一设备包括第一传输接口,所述方法包括:采集第一媒体信息;对所述第一媒体信息进行特征提取,确定所述第一媒体信息的第一特征数据;通过所述第一传输接口向第二设备发送所述第一特征数据,所述第一特征数据用于所述第二设备获得第一应用的结果。
- 如权利要求1所述的方法,其特征在于,所述方法还包括:响应于在所述第一应用上的第一操作,向所述第二设备发送第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;接收所述第二设备返回的第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启所述第一应用协同。
- 如权利要求1或2所述的方法,其特征在于,所述向第二设备发送所述第一特征数据之前,还包括:通过所述第一传输接口接收所述第二设备发送的能力协商请求消息;所述能力协商请求消息用于请求所述第一设备支持的传输协议,及所述第一设备的特征提取能力;所述第一设备的传输协议用于指示所述第一设备支持传输特征数据;所述第一设备的特征提取能力用于指示所述第一设备支持提取所述第一媒体信息的第一特征数据;通过所述第一传输接口向所述第二设备发送能力协商响应消息;所述能力协商响应消息用于确认所述第一设备支持传输特征数据的传输协议及所述第一设备的特征提取能力。
- 如权利要求1-3任一项所述的方法,其特征在于,所述对所述第一媒体信息进行特征提取之前,还包括:获取第一特征提取模型;其中,所述第一特征提取模型用于对所述第一媒体信息进行特征提取,所述第一特征提取模型的版本与第一特征数据处理模型的版本对应,所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果。
- 如权利要求4所述的方法,其特征在于,所述能力协商响应消息还包括:所述第一设备中的特征提取模型的版本;或者,所述第二设备中的特征数据处理模型的版本。
- 如权利要求1-5任一项所述的方法,其特征在于,所述方法还包括:通过所述第一传输接口接收来自所述第二设备的第一消息;所述第一消息用于指示所述第一设备采集媒体信息的状态;响应于所述第一消息,调整所述第一设备采集媒体信息的状态。
- 如权利要求1-6任一项所述的方法,其特征在于,所述方法还包括:通过所述第一传输接口接收来自所述第二设备的第二消息,所述第二消息用于指示所述第一设备采集第三媒体信息的特征数据;响应于所述第二消息,采集所述第三媒体信息;对所述第三媒体信息进行特征提取,得到第三特征数据;通过所述第一传输接口向所述第二设备发送所述第三特征数据。
- 如权利要求6-7任一项所述的方法,其特征在于,所述第二消息或所述第一消息为所述第二设备根据所述第一特征数据确定的。
- 如权利要求1-8任一项所述的方法,其特征在于,所述方法还包括:通过所述第一传输接口接收所述第二设备发送的认证请求消息,所述认证请求消息用于请求所述第一设备是否与所述第二设备建立通信连接,所述通信连接用于确认所述第二设备控制所述第一设备的权限;通过所述第一传输接口向所述第二设备发送认证响应消息;所述认证响应消息用于确认所述第二设备控制所述第一设备的权限;通过所述第一传输接口接收所述第二设备发送的认证成功消息;所述认证成功消息包括:所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识。
- 如权利要求1-9任一项所述的方法,其特征在于,所述第一设备还包括第三传输接口;所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第一设备发送的特征数据或消息为通过所述第一传输接口封装为第一比特流数据后,通过所述第三传输接口发送的。
- 如权利要求1-10任一项所述的方法,其特征在于,所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第一设备接收的消息为通过所述第三传输接口接收的第二比特流数据,并通过所述第一传输接口将所述第二比特流数据解封装后得到的。
- 如权利要求1-11任一项所述的方法,其特征在于,所述第一设备还包括显示单元;所述方法还包括:通过所述第一传输接口接收来自所述第二设备的第三消息,所述第三消息为所述第二设备根据所述第一特征数据确定的所述第一应用的结果,所述第三消息用于指示所述第一设备显示的内容;响应于所述第三消息,通过显示单元显示所述第三消息中用于指示所述第一设备显示的内容。
- 一种媒体信息传输方法,其特征在于,应用于第二设备;所述第二设备包括第二传输接口;所述方法包括:通过所述第二传输接口接收来自第一设备的第一特征数据;所述第一特征数据为根据第一设备对采集的第一媒体信息进行特征提取后确定的;对所述第一特征数据进行处理,得到第一应用的处理结果。
- 如权利要求13所述的方法,其特征在于,所述方法还包括:响应于在所述第一应用上的第二操作,向所述第一设备发送第一通知消息;所述第一设备为与所述第二设备建立通信连接的电子设备;所述第一通知消息用于请求所述第一设备与所述第二设备建立第一应用协同;接收所述第一设备返回的第一响应消息;所述第一响应消息用于确认所述第一设备与所述第二设备开启第一应用协同。
- 如权利要求13或14所述的方法,其特征在于,所述接收来自第一设备的第一特征数据之前,还包括:通过所述第二传输接口向所述第一设备发送能力协商请求消息;所述能力协商请求消 息用于请求所述第一设备支持的传输协议,及所述第一设备的特征提取能力;所述第一设备的传输协议用于指示所述第一设备支持传输特征数据;所述第一设备的特征提取能力用于指示所述第一设备支持提取所述第一媒体信息的第一特征数据;通过所述第二传输接口接收所述第一设备发送的能力协商响应消息;所述能力协商响应消息用于确认所述第一设备支持传输特征数据的传输协议。
- 如权利要求13-15任一项所述的方法,其特征在于,所述接收所述第一特征数据之前,还包括:获取第一特征数据处理模型;所述第一特征数据处理模型用于所述第二设备对所述第一特征数据进行处理获得所述第一应用的结果;第一特征提取模型的版本与所述第一特征数据处理模型的版本对应,所述第一特征提取模型用于对所述第一媒体信息进行特征提取。
- 如权利要求13-16任一项所述的方法,其特征在于,所述方法还包括:通过所述第二传输接口向所述第一设备发送第一消息;所述第一消息用于指示所述第一设备采集媒体信息的状态;所述第一设备采集媒体信息的状态包括以下至少一项:开启状态、关闭状态或采集媒体信息的参数。
- 如权利要求13-17任一项所述的方法,其特征在于,所述方法还包括:通过所述第二传输接口向第一设备发送第二消息;所述第二消息用于指示所述第一设备采集第三媒体信息的特征数据;通过所述第二传输接口接收来自所述第一设备发送的第三特征数据;所述第三特征数据为所述第一设备对采集的第三媒体信息进行特征提取后确定的。
- 如权利要求17-18任一项所述的方法,其特征在于,所述第一消息或所述第二消息为根据所述第一特征数据的处理结果确定的。
- 如权利要求13-19任一项所述的方法,其特征在于,所述第一设备的数量为N个;所述方法还包括:通过所述第二传输接口接收第四消息;所述第四消息包括所述N个第一设备的M个第一特征数据;N、M为大于1的正整数;M大于或等于N;根据所述M个第一特征数据对应的特征数据处理模型,对所述M个第一特征数据进行处理,得到所述第一应用的结果。
- 如权利要求17-20任一项所述的方法,其特征在于,所述方法还包括:通过所述第二传输接口向所述第一设备发送认证请求消息,所述认证请求消息用于请求所述第一设备是否与所述第二设备建立通信连接;所述通信连接用于确认所述第二设备控制所述第一设备的权限;通过所述第二传输接口接收所述第二设备发送的认证响应消息;所述认证响应消息用于确认所述第一设备与所述第二设备建立通信连接。
- 如权利要求21所述的方法,其特征在于,所述方法还包括:响应于所述第二设备发送的认证响应消息,为所述第一设备设置所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布式系统的标识;所述第一设备对应的设备标识及所述分布式系统的标识用于所述第一设备和所述第二设备进行通信;所述第一设备通过所述第一传输接口向所述第二设备发送认证成功消息;所述认证成功消息包括:所述第一设备对应的设备标识,及所述第一设备和所述第二设备所在的分布 式系统的标识。
- 如权利要求13-22任一项所述的方法,其特征在于,所述第二设备还包括第三传输接口;所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第二设备发送的消息为通过所述第二传输接口封装为第二比特流数据后,通过所述第三传输接口发送的。
- 如权利要求13-23任一项所述的方法,其特征在于,所述第一设备与所述第二设备通过第三传输接口建立信道连接;所述第二设备接收的特征数据或消息为通过所述第三传输接口接收的第一比特流数据,并通过所述第二传输接口将所述第二比特流数据解封装后获得的。
- 如权利要求13-24任一项所述的方法,其特征在于,所述第二设备还包括显示单元,所述方法还包括:通过所述显示单元显示所述第一应用的结果。
- 一种电子设备,其特征在于,所述电子设备包括存储器和一个或多个处理器;其中,所述存储器用于存储计算机程序代码,所述计算机程序代码包括计算机指令;当所述计算机指令被所述处理器执行时,使得所述电子设备执行如权利要求1至12中任一项所述的方法。
- 一种电子设备,其特征在于,所述电子设备包括存储器和一个或多个处理器;其中,所述存储器用于存储计算机程序代码,所述计算机程序代码包括计算机指令;当所述计算机指令被所述处理器执行时,使得所述电子设备执行如权利要求13至25中任一项所述的方法。
- 一种媒体信息传输系统,其特征在于,包括:如权利要求26所述的电子设备和如权利要求27所述的电子设备。
- 一种计算机可读存储介质,其特征在于,所述计算机可读存储介质包括程序指令,当所述程序指令在电子设备上运行时,使得所述电子设备执行如权利要求1至12任一项所述的方法,或者,使得所述电子设备执行如权利要求13至25任一项所述的方法。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP21855430.1A EP4187421A4 (en) | 2020-08-14 | 2021-08-04 | MEDIA INFORMATION TRANSMISSION METHOD AND ELECTRONIC DEVICE |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010819860 | 2020-08-14 | ||
CN202010819860.3 | 2020-08-14 | ||
CN202011300677.9 | 2020-11-19 | ||
CN202011300677.9A CN114077747A (zh) | 2020-08-14 | 2020-11-19 | 一种媒体信息传输方法及装置 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022033377A1 true WO2022033377A1 (zh) | 2022-02-17 |
Family
ID=80247689
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2021/110644 WO2022033377A1 (zh) | 2020-08-14 | 2021-08-04 | 一种媒体信息传输方法及电子设备 |
Country Status (2)
Country | Link |
---|---|
EP (1) | EP4187421A4 (zh) |
WO (1) | WO2022033377A1 (zh) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115204204A (zh) * | 2022-09-19 | 2022-10-18 | 江苏省质量和标准化研究院 | 基于mt-bp电磁感知的uhf多标签自适应识读方法及系统 |
CN115278591A (zh) * | 2022-05-19 | 2022-11-01 | 深圳市航盛电子股份有限公司 | 投屏传输方法、装置、终端设备以及存储介质 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104123930A (zh) * | 2013-04-27 | 2014-10-29 | 华为技术有限公司 | 喉音识别方法及装置 |
CN104378589A (zh) * | 2014-11-11 | 2015-02-25 | 深圳市视晶无线技术有限公司 | 一种低码率视频编码传输重构方法及系统 |
CN104883344A (zh) * | 2014-02-28 | 2015-09-02 | 华为技术有限公司 | 协商媒体能力的方法和设备 |
CN110430068A (zh) * | 2018-04-28 | 2019-11-08 | 华为技术有限公司 | 一种特征工程编排方法及装置 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10970605B2 (en) * | 2017-01-03 | 2021-04-06 | Samsung Electronics Co., Ltd. | Electronic apparatus and method of operating the same |
CN110675421B (zh) * | 2019-08-30 | 2022-03-15 | 电子科技大学 | 基于少量标注框的深度图像协同分割方法 |
-
2021
- 2021-08-04 EP EP21855430.1A patent/EP4187421A4/en active Pending
- 2021-08-04 WO PCT/CN2021/110644 patent/WO2022033377A1/zh unknown
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104123930A (zh) * | 2013-04-27 | 2014-10-29 | 华为技术有限公司 | 喉音识别方法及装置 |
CN104883344A (zh) * | 2014-02-28 | 2015-09-02 | 华为技术有限公司 | 协商媒体能力的方法和设备 |
CN104378589A (zh) * | 2014-11-11 | 2015-02-25 | 深圳市视晶无线技术有限公司 | 一种低码率视频编码传输重构方法及系统 |
CN110430068A (zh) * | 2018-04-28 | 2019-11-08 | 华为技术有限公司 | 一种特征工程编排方法及装置 |
Non-Patent Citations (1)
Title |
---|
See also references of EP4187421A4 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115278591A (zh) * | 2022-05-19 | 2022-11-01 | 深圳市航盛电子股份有限公司 | 投屏传输方法、装置、终端设备以及存储介质 |
CN115204204A (zh) * | 2022-09-19 | 2022-10-18 | 江苏省质量和标准化研究院 | 基于mt-bp电磁感知的uhf多标签自适应识读方法及系统 |
Also Published As
Publication number | Publication date |
---|---|
EP4187421A4 (en) | 2023-12-27 |
EP4187421A1 (en) | 2023-05-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021051989A1 (zh) | 一种视频通话的方法及电子设备 | |
JP7378576B2 (ja) | リモコン装置を用いたワンタッチスクリーンプロジェクションを実施するための端末装置、方法及びシステム | |
CN112449332B (zh) | 一种蓝牙连接的方法和电子设备 | |
WO2021051986A1 (zh) | 设备之间建立连接的方法及电子设备 | |
CN111316598A (zh) | 一种多屏互动方法及设备 | |
WO2022100304A1 (zh) | 应用内容跨设备流转方法与装置、电子设备 | |
WO2022121775A1 (zh) | 一种投屏方法及设备 | |
WO2022033377A1 (zh) | 一种媒体信息传输方法及电子设备 | |
CN110166820A (zh) | 音视频的播放方法、终端及装置 | |
CN115334138B (zh) | Quic数据传输方法、装置、客户端及服务端 | |
WO2021104114A1 (zh) | 一种提供无线保真WiFi网络接入服务的方法及电子设备 | |
EP4283931A1 (en) | Nfc method and system, and electronic device | |
WO2024087900A1 (zh) | 一种摄像头切换方法及相关电子设备 | |
WO2022135163A1 (zh) | 一种投屏显示方法及电子设备 | |
WO2022166618A1 (zh) | 一种投屏的方法和电子设备 | |
US11895713B2 (en) | Data sharing and instruction operation control method and system | |
CN113676879A (zh) | 一种分享信息的方法、电子设备和系统 | |
CN113408016A (zh) | 保存密文的方法和装置 | |
WO2022033379A1 (zh) | 一种媒体信息传输方法及装置 | |
WO2021037208A1 (zh) | 一种蓝牙连接的方法和电子设备 | |
CN113946302B (zh) | 一种打开文件的方法及设备 | |
EP4394578A1 (en) | Screen mirroring data processing method, electronic device, and storage medium | |
CN114077747A (zh) | 一种媒体信息传输方法及装置 | |
WO2022174718A1 (zh) | 一种数据备份方法和电子设备 | |
WO2022007750A1 (zh) | 屏幕共享方法、终端和存储介质 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21855430 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2021855430 Country of ref document: EP Effective date: 20230221 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |