WO2023231894A1 - 基于协同纠错的唤醒方法、装置及系统、介质、设备 - Google Patents

基于协同纠错的唤醒方法、装置及系统、介质、设备 Download PDF

Info

Publication number
WO2023231894A1
WO2023231894A1 PCT/CN2023/096289 CN2023096289W WO2023231894A1 WO 2023231894 A1 WO2023231894 A1 WO 2023231894A1 CN 2023096289 W CN2023096289 W CN 2023096289W WO 2023231894 A1 WO2023231894 A1 WO 2023231894A1
Authority
WO
WIPO (PCT)
Prior art keywords
voice
wake
voice device
devices
request
Prior art date
Application number
PCT/CN2023/096289
Other languages
English (en)
French (fr)
Inventor
陈峰峰
张新星
高向军
邓宏
袁伟
Original Assignee
四川虹美智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 四川虹美智能科技有限公司 filed Critical 四川虹美智能科技有限公司
Publication of WO2023231894A1 publication Critical patent/WO2023231894A1/zh

Links

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/28—Constructional details of speech recognition systems
    • G10L15/34—Adaptation of a single recogniser for parallel processing, e.g. by use of multiple processors or cloud computing
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00—Data switching networks
    • H04L12/28—Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/2803—Home automation networks
    • H04L12/2816—Controlling appliance services of a home automation network by calling their functionalities
    • H04L12/282—Controlling appliance services of a home automation network by calling their functionalities based on user interaction within the home
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
    • G10L2015/225—Feedback of the input speech
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Definitions

  • One or more embodiments of this specification relate to the technical field of voice equipment, and in particular, a wake-up method, device, system, medium, and equipment based on collaborative error correction.
  • Speech recognition technology is currently the most widely used human-computer interaction technology. Users initiate voice commands to control voice devices to perform corresponding operations, such as turning on or off the air conditioner.
  • the whole process of voice interactive control includes several key links: voice wake-up, sending voice commands, device execution actions, device voice feedback, etc.
  • Voice wake-up means pre-setting one or more proprietary words for the voice device. The user can use this proprietary vocabulary to activate the voice device in the wake-up waiting state and enter the voice command recognition waiting state. The user can further initiate various voice commands. , after receiving it, the voice device executes the corresponding command action, and uses the device's own playback component to feedback the result of the command execution and prompt the user.
  • the voice device When the user speaks proprietary words, the voice device can be activated. When the user speaks non-proprietary words, the voice device remains awake and waiting to avoid the user's normal interpersonal voice communication from accidentally triggering the voice. However, in actual scenarios, there are still situations where voice devices are awakened and activated by non-proprietary words or even environmental noise. The industry calls this type of awakening a false awakening, and the probability index generally ranges from 1-3 times/72 hours. False triggering will disturb users and reduce user experience.
  • One or more embodiments of this specification describe a wake-up method, device, system, medium, and equipment based on collaborative error correction.
  • this specification provides a wake-up method based on collaborative error correction.
  • Voice devices are distributed throughout the entire space of a home.
  • the voice devices include smart home appliances.
  • the voice modules in the smart home appliances have voice pickup capabilities. and a voice feedback function; each of the voice devices is connected to the cloud platform; the method is executed by a decision-making device, and the decision-making device is a smart home appliance selected in advance from the voice devices; the method includes:
  • a voice device in a wake-up waiting state monitors the wake-up word and sends a wake-up message to the decision-making device. Request that all voice devices in the entire home space have the same wake word;
  • start timing and after the timing reaches the preset duration, determine whether a wake-up request sent by the other voice devices is received during the timing process;
  • a responding voice device is determined from the voice device that sent the wake-up request, and a wake-up response instruction is sent to the responding voice device, so that the responding voice device The device enters the wake state from the wake wait state.
  • this specification provides a wake-up device based on collaborative error correction.
  • Voice devices are distributed throughout the entire space of a home.
  • the voice devices include smart home appliances.
  • the voice modules in the smart home appliances have voice pickup capabilities. and voice feedback function; each voice device is connected to the cloud platform;
  • the device is installed on a decision-making device, which is a smart home appliance selected in advance from the voice device; the device includes:
  • the first judgment module is used to judge whether there are other voice devices in the space group where the voice device is located when receiving a wake-up request sent by a voice device; wherein a voice device in a wake-up waiting state monitors the wake-up word. Send a wake-up request to the decision-making device, and each voice device in the entire space of a home has the same wake-up word;
  • the second judgment module is used to start timing if there are other voice devices, and determine whether a wake-up request sent by other voice devices is received during the timing process after the timing reaches the preset length;
  • the target determination module is used to determine if a wake-up request from other voice devices is received during the timing process.
  • a responding voice device is determined among the voice devices that send the wake-up request, and a wake-up response instruction is sent to the responding voice device, so that the responding voice device enters the awakening state from the wake-up waiting state.
  • embodiments of the present invention provide a wake-up system based on collaborative error correction, which includes voice devices distributed in a home and a cloud platform that communicates and connects with each voice device; the voice devices include smart home appliances, so The voice module in the smart home appliance has voice pickup and voice feedback functions; the cloud platform is used to control voice devices in the entire space; the decision-making device is a smart home appliance selected in advance from the voice device, The decision-making device has a wake-up device based on collaborative error correction provided in the second aspect.
  • embodiments of the present invention provide a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, the steps of the method provided in the first aspect are implemented.
  • embodiments of the present invention provide a voice device, including a memory, a processor, and a computer program stored in the memory and executable on the processor.
  • the processor executes the computer program, the first aspect is implemented. The steps of the method are provided.
  • embodiments of the present invention provide a complete set of home appliance control system, including an Internet of Things platform, a content platform, a semantic platform, a cloud platform, a voice acquisition module, and an intelligent terminal; wherein the Internet of Things platform is the Zhihuijia AIOT platform, and the The Internet of Things platform has a configuration module and an intelligent control scene module.
  • the configuration module and the intelligent control scene module transmit information with intelligent terminals through the communication network and control the operation of the intelligent terminal;
  • the cloud platform communicates with the semantic platform and the Internet of Things platform through the communication network respectively.
  • Transmission; the semantic platform includes a speech analysis module, a functional feedback module and an interactive feedback module.
  • the semantic platform can analyze the speech signal obtained by the speech acquisition module and output control instructions based on the analyzed speech signals.
  • the control instructions are communicated with the cloud through wireless communication.
  • the platform communicates with the Internet of Things platform to control or drive the operation of smart terminals.
  • the decision-making device when the decision-making device receives a wake-up request sent by a voice device, the decision-making device The device will determine whether there are other voice devices in the space group where the voice device that sent the wake-up request is located. If there are other voice devices, it will start timing. After the time is over, it will determine whether it has received wake-up calls from other voice devices during the timing process.
  • the decision-making device receives wake-up requests from multiple voice devices. At this time, the decision-making device will select the most appropriate voice device as the responding voice device, which can provide users with a high-quality user experience.
  • Figure 1 is a schematic flowchart of a wake-up method based on collaborative error correction in one embodiment of this specification
  • Figure 2 is a schematic diagram of four voice devices accessing the same network in an embodiment of this specification
  • Figure 3 is a schematic diagram of room allocation of four voice devices in one embodiment of this specification.
  • Figure 4 is a schematic flowchart of a voice wake-up processing method in one embodiment of this specification
  • Figure 5 is a schematic diagram of the distribution of multiple smart home appliances in a home according to an embodiment of this specification
  • Figure 6a is a schematic diagram of a voice terminal in an embodiment of this specification.
  • Figure 6b is a schematic diagram of the distribution of multiple voice terminals in a home in an embodiment of this specification
  • Figure 7 is a structural block diagram of a wake-up device based on collaborative error correction in one embodiment of this specification
  • Figure 8 is a structural block diagram of a wake-up system based on collaborative error correction in an embodiment of this specification.
  • Figure 9 is an architectural diagram of a complete home appliance system in one embodiment of this specification.
  • Figure 10 is a schematic diagram of the control principle of the voice command system in one embodiment of this specification.
  • embodiments of the present invention provide a wake-up method based on collaborative error correction.
  • voice devices distributed in a home. These voice devices include various smart home appliances. For example, see Figure 5. There is a washing machine, a refrigerator, two hanging air conditioners, and a cabinet-type air conditioner in the entire space of a home. Air conditioning, two TVs. These smart home appliances all have voice modules.
  • the voice modules in smart home appliances can pick up voice commands issued by users and send the voice commands to the cloud platform.
  • the voice module can also provide voice feedback under the control of the cloud platform, that is, inform the user of the execution status of some devices through voice broadcast.
  • voice terminals are installed in the entrance area, balcony, dining room, study room, secondary bathroom, main bathroom, bedside of the master bedroom, bedside of the children's room, etc.
  • the voice terminal can pick up the voice uttered by the user and can also access the home LAN through WIFI. Since the size of the voice terminal is very small and it is inconvenient to install a loudspeaker module, the voice terminal does not have the function of voice broadcasting.
  • the embodiment of the present invention is Voice terminals are set up in areas within the home that are not covered by smart home appliances. In this way, the optimal sound pickup range of each smart home appliance and each voice terminal can cover the entire space in a home.
  • the voice terminal can also have a light prompt function. For example, when a voice terminal is awakened as a voice-responsive device, it can be prompted by a breathing light.
  • the cloud platform can control the corresponding target smart home appliance as a voice-responsive device after successfully executing the command. The voice terminal uses a green light to flash three times quickly to prompt.
  • the cloud platform controls the corresponding target smart home appliance and fails to execute the command, it can control the voice terminal as the responding voice device to use a yellow light to flash three times quickly to prompt.
  • all voice devices in a home are connected to the cloud platform, so that the cloud platform can control each voice device in the home.
  • an application can be installed on the user's mobile terminal, and the application can also control and configure various voice terminals in the home.
  • the user utters the wake-up word "Changhong Xiaobai" on the balcony.
  • the voice terminal on the balcony, the TV in the living room, and the cabinet air conditioner in the living room will all pick up the wake-up word. So which voice device should be awakened? This is Problems to be solved by embodiments of the invention. If it is determined that the TV in the living room is woken up, the TV in the living room will respond "Yes", and then the user will issue a voice command "Turn on the air conditioner in the living room". At this time, the TV in the living room will pick up the voice and send the voice to the cloud.
  • the platform and the cloud platform analyze the speech and then control the cabinet air conditioner in the living room to turn on.
  • the awakened voice device in the embodiment of the present invention is called a response voice device.
  • the function of the response voice device is to pick up the voice instructions issued by the user and send the voice instructions to the cloud platform, so that the cloud platform will analyze the voice instructions. , control the execution device to perform corresponding operations based on the analysis results.
  • the voice devices include smart home appliances and voice terminals.
  • the optimal sound pickup ranges of each of the smart home appliances and each of the voice terminals are The union can cover the entire space; the voice module in the smart home appliance has the function of voice pickup and voice feedback, and the voice module of the voice terminal has the function of voice pickup; the voice devices are all used for performing full-space operations. Cloud platform communication connection controlled by internal voice device.
  • Embodiments of the present invention provide a wake-up method based on collaborative error correction, which method can be executed by a decision-making device, which is a smart home appliance selected in advance from the voice device.
  • a decision-making device which is a smart home appliance selected in advance from the voice device.
  • this method is applicable to scenarios where a home only includes smart home appliances.
  • the voice module in the smart home appliances has voice pickup and voice feedback functions.
  • voice devices there are voice devices distributed throughout the entire space of a home.
  • the voice devices include smart home appliances and voice terminals, and the union of the optimal sound pickup ranges of each of the smart home appliances and each of the voice terminals. Can cover the entire space; the voice module in the smart home appliance has voice pickup and voice feedback functions, and the voice module of the voice terminal has the voice pickup function; the voice devices are all used to control voice devices in the entire space cloud platform communication connection.
  • a home contains at least two smart home appliances and at least two voice terminals.
  • the applicable scenario for such specific embodiments may be the scenario described in the previous paragraph, or of course it may also be the scenario: in a family
  • This scenario only includes smart home appliances.
  • the voice module in the smart home appliance has voice pickup and voice feedback functions.
  • the method includes the following steps S10 to S30:
  • a voice device in a wake-up waiting state sends a wake-up request to the decision-making device after listening to the wake-up word.
  • Each voice device in the entire space of a home has the same wake-up word.
  • the method provided by the embodiment of the present invention is executed by the decision-making device, and the decision-making device is a smart home appliance among the voice devices in a home. Therefore, before executing the method provided by the embodiment of the present invention, it is necessary to first determine Execution device.
  • each voice device in a family can be controlled through an application on the mobile terminal, and the application is the default application mentioned below. Determination of decision-making equipment can be achieved by this preset application.
  • each voice device is controlled by the same preset application on the user's mobile terminal in the home; the decision-making device is predetermined by the preset application; wherein the preset application determines the
  • the process of decision-making equipment includes the following steps S01 ⁇ S04:
  • the preset application can obtain a device list formed by relevant information of each voice device connected to the home network.
  • the voice device After the voice device has been running for a period of time, a lot of data will be generated, such as the user's usage habits of the voice device and the charging status data of the voice device. These data will be recorded in the memory inside the voice device.
  • the default application will obtain this data from each voice device. At the same time, the default application will also obtain the computing power data of each voice device.
  • the charging status data of the voice device refers to whether the voice device is always powered, charged at a certain time of the day, or rarely charged, etc.
  • the computing power data of the voice device refers to the data that can reflect the computing speed and computing volume of the voice device. Because the decision-making device needs to perform wake-up decision-making processing while realizing its own functions, the computing power of the voice device needs to be taken into consideration here.
  • S03. Determine the decision-making ability score of each voice device based on the charging status data, computing power data and user usage data of each voice device;
  • a first calculation formula may be used to calculate the decision-making ability score of each voice device in S03.
  • P1 is the score of the decision-making ability
  • d is the average daily charging time of the voice device in the past month
  • y is the CPU computing power of the voice device
  • u is the usage mark of the voice device by home users in the current season; if If the home user uses the voice device in the current season, the corresponding usage mark of the voice device is 1; if the home user does not use the voice device in the current season, the corresponding use mark of the voice device is 0.
  • d is the average daily charging time of the voice device in the past month. This parameter not only takes into account the user's usage in the recent period, but also reflects the average charging situation every day. For example, users only turned on voice TV at night in the past month, so the average duration of voice TV in the past month was only a few hours.
  • the voice refrigerator has always been powered on. At this time, the voice refrigerator has been powered on for an average of 24 hours a day in the past month.
  • Parameter d is a key parameter. The longer the voice device is powered on, the more time the voice device can make wake-up decision processing, which can reduce the occurrence of missed wake-up requests. The larger d is, the larger P1 is.
  • y reflects the computing power of the voice device, for example, the computing power of the CPU.
  • the computing power of different CPUs is calculated in different ways. For example, some CPUs measure computing power by word length, and some CPUs measure computing power by double-precision floating point computing capabilities. The specifics can be determined according to the actual situation of the CPU. The larger y is, the larger P1 is, but the relationship between y and P1 is not proportional.
  • the decision-making ability score of each voice device connected to the home network is calculated, and then the voice device with the highest decision-making ability score is selected as the decision-making device.
  • the decision-making device can be used to execute the method provided by the embodiment of the present invention.
  • the decision-making device loses power, other voice devices need to be replaced as the decision-making device.
  • the previously determined decision-making device is not powered, and in this case it is also necessary to replace other voice devices as the decision-making device.
  • the process of the preset application determining the decision-making device may also include: the preset application detects the current decision-making device at a preset time interval through a heartbeat instruction. Whether it is in the charged state; if it is not in the charged state, the voice device with the highest decision-making ability score among the various voice devices in the charged state is selected as the current decision-making device.
  • the preset application on the mobile terminal will detect whether the current decision-making device is charged according to time intervals, specifically through the heartbeat command. For example, the preset application on the mobile terminal will check whether the current decision-making device is charged at intervals. The device sends a heartbeat command. If the mobile terminal can receive feedback information from the decision-making device, it means that the decision-making device is powered on at this time. If it cannot receive feedback information from the decision-making device, it means that the decision-making device is not powered on at this time.
  • the voice device with the highest decision-making ability score will be selected from among the powered voice devices as the current decision-making device. , this can avoid affecting the wake-up processing because the decision-making device is not powered.
  • the preset application on the mobile terminal can also mark the room where each voice device is located on the preset application, that is, setting a corresponding room identification for each voice device. .
  • the above space grouping may be a physical space grouping or a virtual space grouping.
  • one room corresponds to one space grouping
  • one space grouping includes at least one voice device. That is to say, the space grouping is a plurality of groups obtained by dividing the voice devices in each room according to the physical space grouping method.
  • individual devices can be grouped on preset applications.
  • the preset application can be used to provide a configuration interface so that the user configures a corresponding room identification for each voice device connected to the home network; correspondingly, as described in S10, determine the space group in which the voice device is located.
  • Whether there are other voice devices in the room includes: obtaining the room ID corresponding to the voice device from the preset application program, using the room ID as the first room ID; determining whether there are room IDs of other voice devices that match the first room ID. A room with the same room ID; if it exists, there are other voice devices in the room where the voice device is located; otherwise, there are no other voice devices in the room where the voice device is located.
  • the user can set a corresponding room identification for each voice device on the configuration interface of the default application, so that the default application can know which voice devices are in each room.
  • the decision-making device determines whether there are other voice devices in the space group where the voice device is located, it will obtain the room identifier corresponding to the voice device that sends the wake-up request, that is, the first room identifier, from the preset application program of the mobile terminal. Then it is determined whether the room identifiers of other voice devices have the same room identifier as the first room identifier, and then it is determined whether there are other voice devices in the room where the voice device that sent the wake-up request is located.
  • a voice device in a room that is in a wake-up waiting state listens to the wake-up word, and then sends a wake-up request to the decision-making device, and then the decision-making device determines the room where the voice device is located. Check whether there are other voice devices, and then perform the subsequent steps.
  • the processing method is similar.
  • the decision-making device will start timing at this time.
  • the timing stops after the preset duration is reached.
  • a wake-up request from other voice devices may be received, or a wake-up request from other voice devices may not be received.
  • determining a responding voice device from the voice device that sent the wake-up request includes:
  • the voice pickup parameters of each voice device that sends a wake-up request are obtained;
  • the responding voice device is determined from the voice devices that sent the wake-up request according to the sound pickup parameters of each voice device that sent the wake-up request.
  • a wake-up request is received from other voice devices, it means that more than one voice device has responded to the user's voice. At this time, the possibility of false wake-up will be relatively small, so at this time, it is necessary to Select the wake-up object with the highest probability from each voice device that issues a wake-up request. If no wake-up request from other voice devices is received during the timing process, there is a high possibility that the voice device that issued the wake-up request will be woken up by mistake.
  • the pickup parameters of each wake-up request can be obtained, and then the pickup parameters can be used for the next step of calculation.
  • the so-called voice pickup parameters refer to some parameters when the voice device collects the voice uttered by the user.
  • the sound pickup parameters may include sound pickup angle and sound pickup intensity. If the user speaks to the voice device that he wants to wake up, the sound pickup angle at this time is 90°, but the user does not speak to the voice device that he wants to wake up. When the user speaks, the sound pickup angle is not 90°. It can be seen that the sound pickup angle reflects the sound pickup angle of the voice device when the user speaks. Pickup sound intensity refers to the intensity of the sound picked up by the voice device.
  • the responding voice device is determined according to the respective sound pickup parameters of each voice device that sends the wake-up request.
  • the specific process may include the following steps S41 to S43:
  • 60 to 120 degrees is a better sound pickup angle.
  • the sound pickup angle is 60 to 120 degrees, the possibility of the voice device responding to the voice device is very high.
  • the number of voice devices with sound pickup angles in the range of 60 to 120 degrees is first determined. If the number is greater than 1, then each of these voice pickup angles in the range of 60 to 120 degrees needs to be installed. Further filtering in voice devices.
  • the number of voice devices with pickup angles in the range of 60 to 120 degrees is greater than 1, calculate the voice based on the pickup angle and pickup sound intensity of each voice device with a pickup angle in the range of 60 to 120 degrees.
  • the arousal score corresponding to the device; the arousal score of a voice device is used to represent the probability of the user waking up the voice device;
  • the wake-up score of each voice device is calculated based on the pickup angle and sound intensity of the voice device, so as to know that the user wants to wake up the voice device. What is the probability of a voice device.
  • a second calculation formula may be used to calculate the arousal score corresponding to the voice device.
  • the second calculation formula includes:
  • P2 is the arousal score
  • s is the sound pickup intensity
  • r is the pickup angle
  • the first range is: r is greater than or equal to 60 and less than 80, or r is greater than 100 and less than or equal to 120.
  • the second range is: r is greater than or equal to 80 and less than or equal to 100; a and b are preset weights, and a is greater than b, and c is 10*a+b/10.
  • the weight of the pickup sound intensity is a, and a is greater than b, which means that more attention is paid to the pickup sound intensity at this time. In other words, the pickup sound intensity is more important in this case.
  • the weight of the pickup angle is a, and a is greater than b, which means that more attention is paid to the pickup angle at this time. That is to say, in this case, the pickup angle is more important.
  • the wakeup score should be greater than the wakeup score when r is within the first range.
  • the wakeup score when r is within the second range The parameter c is added to the calculation formula, and c is 10*a+b/10. This value can ensure that when the pickup sound intensity is the same, the arousal score of r in the second range is greater than the arousal score of r in the first range. score.
  • the arousal score of r in the second range is greater than the arousal score of r in the first range under the same pickup sound intensity. score.
  • the higher the wake-up score of a voice device the greater the probability that the user wants to wake up the voice device.
  • the lower the wake-up score of a voice device the lower the probability that the user wants to wake up the voice device. Therefore, the voice device with the highest wake-up score will be used as the answering voice device.
  • the responding voice device is determined based on the sound pickup parameters of each voice device that sends the wake-up request.
  • the specific process may also include at least one of the following:
  • this voice device is used as the answering voice device at this time.
  • the number of voice devices with pickup angles in the range of 60 to 120 degrees is 0, it means that the user has not spoken to any voice device. At this time, there is no need to consider the pickup angle. At this time, you can issue a wake-up call. Among the requested voice devices, the voice device with the strongest pickup sound is selected as the responding voice device.
  • the answering voice device can be determined according to the above process in different situations.
  • the above are all processing methods when a wake-up request sent by other voice devices is also received during the timing process.
  • the voice device that sent the wake-up request can be considered as a false trigger.
  • the decision-making device can send a termination wake-up response instruction to the voice device that sent the wake-up request.
  • the voice device that receives the termination wake-up response instruction will not enter the wake-up state from the wake-up waiting state, but will continue to remain in the wake-up waiting state.
  • the method provided by the embodiment of the present invention may also include the following steps: sending a wake-up response instruction to the voice device that sends the wake-up request, so that the voice device enters the wake-up state. .
  • the voice device sends a wake-up request to the decision-making device.
  • the decision-making device thinks that the user wants to wake up the voice device, so the decision-making device will send a wake-up response to the device. command, the voice device that receives the command will enter the wake-up state from the wake-up waiting state.
  • the decision-making device in the embodiment of the present invention may be in the same room as the voice device that issues the wake-up request, or may not be in the same spatial group as the voice device that issues the wake-up request.
  • each voice device in the embodiment of the present invention is connected to the same network, for example, connected to the same home network through Ethernet, Bluetooth, WIFI, etc. In this way, various voice devices can communicate with each other and transfer data,
  • Voice device A There are voice device A, voice device B, voice device C, and voice device D in a home.
  • Voice device A and voice device B use WIFI network technology, and voice device C uses wired Ethernet technology.
  • Voice device D uses Bluetooth communication technology and is connected to home network devices, and voice device A is selected as the decision-making device.
  • the user marks the room identification of each voice device in the application of his mobile phone.
  • voice device A, voice device B, and voice device C are assigned to room A through the APP software
  • voice device D is assigned to In room B, deliver the device list and device ID information of room A and room B to voice device A.
  • voice device B when voice device B is triggered by the user's voice and needs to wake up, it first initiates a wake-up request to the decision-making device, namely voice device A, through the voice device B network.
  • the decision-making device checks that there is still voice in the room A group of voice device B.
  • Device A and voice device C enter a timer wait. After the timer expires, it is determined whether voice device A and voice device C have also initiated a wake-up request. It is judged that only voice device B submitted the wake-up request this time, so it is judged as a false wake-up this time, and the decision-making device sends a termination response instruction to voice device B.
  • the method provided by the embodiment of the present invention is to solve the problem that the probability of false triggering increases exponentially when there are multiple voice devices in a home, and proposes a voice wake-up solution that supports collaborative error correction.
  • this solution It is necessary to first connect the device to the network, select the decision-making device, and mark the room, and then execute the above-mentioned steps S1 to S4 provided by the embodiment of the present invention.
  • the space grouping may also be a virtual space grouping.
  • the space grouping is a virtual space grouping divided according to the position and orientation of the user who issued the wake-up word in the whole space.
  • One virtual space grouping includes at least two voice devices; the space grouping is the cloud
  • the platform is predetermined.
  • the virtual space grouping is not set by the user on the application, but is divided by the cloud platform based on the user's location and orientation, and no user setting is required.
  • the cloud platform For virtual space grouping, the cloud platform’s methods for determining space grouping include:
  • a2 Obtain the sound intensity of the preset wake-up word picked up by each voice device, and select the first voice device from each voice device according to the sound intensity; wherein, the sound picked up by each first voice device The intensity is higher than the sound intensity picked up by other voice devices, and the maximum difference between the sound intensities picked up by each of the first voice devices is within the preset range;
  • each of the second voice devices into a spatial grouping for the user's position and orientation.
  • the sound intensity picked up by the five voice devices at the entrance, living room and dining room is slightly higher than that in the laundry room and kitchen.
  • the sound intensity picked up by the voice devices, the sound intensity picked up by the five voice devices in the home, living room and dining room is much higher than the sound intensity picked up by the voice devices in the study, bedroom and bathroom.
  • the first voice device is selected according to the sound intensity picked up by each voice device. Since the user is standing in the middle of the living room and dining room, the sound intensity picked up by the five voice devices in the home, living room and dining room is almost the same, and these five voice devices have the strongest sound intensity among all voice devices. , so these 5 voice devices are regarded as the first voice devices.
  • the three voice devices located in the living room and balcony face the direction of the user's voice, while the two voice devices at the entrance and the restaurant face away from the direction of the user's voice, so the three voice devices in the living room and balcony face the direction of the user's voice.
  • the pickup angles of the three voice devices are within the optimal pickup angle range, but the pickup angles of the two voice devices at the entrance and the restaurant are not within the optimal pickup angle range. Therefore, the three voice devices in the living room and balcony are The device serves as a second voice device, and these three second voice devices form a virtual space group.
  • the two voice devices in the restaurant and at home form a virtual space group.
  • a virtual space group can be corresponding to the user's position and orientation. In fact, when the user is at a certain location, a virtual space group can be corresponding. For example, when the user is sitting on the sofa in the living room and watching TV, the corresponding virtual space at this time A virtual space grouping formed by three voice devices grouped into the living room and balcony.
  • multiple virtual space groups can be formed for multiple wake-up requests, and the multiple virtual space groups can be saved, and the virtual space groups can be used directly in the future. .
  • Virtual space grouping is a virtual space grouping automatically divided based on past experience.
  • a virtual space group is a device group determined based on the user's location and orientation.
  • Each voice device in a virtual space group can be located in the same physical space or in multiple physical spaces. For example, when the user stands in the middle of the living room and dining room and makes a wake-up request toward the balcony, the three devices, the cabinet air conditioner and TV in the living room, and the voice terminal on the balcony will be divided into a virtual space group.
  • a virtual space group is a virtual space group. There are at least two voice devices in this virtual space group. If there is only one voice device, there is no need to divide it into a group.
  • the decision-making device when it receives a wake-up request sent by a voice device, it determines whether there are other voice devices in the virtual space group where the voice device is located. If there are other voice devices, it is determined whether the wake-up request sent by the other voice devices can be received within a certain period of time. If a wake-up request sent from another voice device is received, it means that the possibility of false wake-up is relatively small, and then one voice device is selected as the responding voice device among the voice devices that send the wake-up request. However, if there are other voice devices in the virtual space group, but no wake-up request from other voice devices is received within a certain period of time, it means that the possibility of false wake-up is relatively high, and the response voice is uncertain at this time. equipment.
  • a voice device does not have a corresponding virtual space group. For example, there is only one voice device in a room.
  • the voice device is not divided into any virtual space group.
  • the decision-making device receives the wake-up request sent by the voice device, it will directly use the voice device as the responding voice device.
  • each voice terminal is connected to the cloud platform, which can form a control for each voice terminal in the entire space of a home.
  • the method provided by the embodiment of the present invention may be a nearby wake-up method for collaborative error correction.
  • the steps of the nearby wake-up method are consistent with the above-mentioned steps S10 to S30.
  • a responding voice device is determined from the voice devices that sent the wake-up request based on the proximity principle.
  • the so-called proximity principle selects the voice device closest to the user from the voice devices that send the wake-up request as the responding voice device.
  • a voice device closest to the user is selected as the answering voice device among multiple voice devices that meet the sound intensity requirements and pickup angle range requirements.
  • the decision-making device when the user issues a wake-up word such as "Changhong Xiaobai” or "Haier Xiaohei", the decision-making device can also be directly awakened, and the decision-making device can wake up and control other voice devices.
  • a wake-up word such as "Changhong Xiaobai” or "Haier Xiaohei”
  • each voice device in a family may also have different wake-up words, but the above method is performed for each voice device with the same wake-up word.
  • the voice terminal does not have the voice feedback function. If a speaker module is integrated into the voice terminal so that the voice terminal has a voice feedback function, the voice terminal can be regarded as a smart home appliance, because generally the voice terminal has the function of a reminder light. At this time, the voice terminal can be regarded as a smart home appliance. It is a smart home appliance with a reminder light function.
  • the voice device can communicate directly with the cloud platform, or it can communicate with the cloud platform through the home local area network.
  • the pickup angle range is 60 to 120 degrees, which can also be expanded to 50 to 130, 40 to 140, 30 to 150, 20 to 160, 10 to 170, etc.
  • the decision-making device when the decision-making device receives a wake-up request sent by a voice device, the decision-making device will determine whether there are other voices in the space group where the voice device that sent the wake-up request is located. The voice device, if there are other voice devices, starts timing. After the timer ends, it determines whether a wake-up request from other voice devices is received during the timing process. If a wake-up request from other voice devices is received, a responding voice device is determined. Then, a wake-up response instruction is sent to the answering voice device, so that the answering voice device will enter the wake-up state from the wake-up waiting state.
  • the decision-making device receives wake-up requests from multiple voice devices. At this time, the decision-making device will select the most suitable voice device as the responding voice device based on the sound pickup parameters. While realizing nearby wake-up, it can provide the user with Bring high-quality user experience.
  • embodiments of the present invention provide a wake-up device based on collaborative error correction.
  • Voice devices are distributed throughout the entire space of a home.
  • the voice devices include smart home appliances, and the voice modules in the smart home appliances have voice functions. Pickup and voice feedback functions; each of the voice devices is connected to the cloud platform; the device is installed on a decision-making device, and the decision-making device is a smart home appliance selected in advance from the voice device; see Figure 7,
  • the device includes:
  • the first judgment module is used to judge whether there are other voice devices in the space group where the voice device is located when receiving a wake-up request from a voice device; wherein, a voice device in a wake-up waiting state monitors After hearing the wake-up word, send a wake-up request to the decision-making device.
  • Each voice device in the entire space of a home has the same wake-up word;
  • the second judgment module is used to start timing if there are other voice devices, and determine whether a wake-up request sent by other voice devices is received during the timing process after the timing reaches the preset length;
  • a target determination module configured to determine a responding voice device from the voice device that sent the wake-up request if a wake-up request sent from other voice devices is received during the timing process, and send a wake-up response instruction to the responding voice device, So that the answering voice device enters the wake-up state from the wake-up waiting state.
  • the sound pickup parameters include pickup angle and sound intensity;
  • the target determination module specifically includes:
  • the first judgment unit is used to: for each voice device that sends a wake-up request, determine whether the number of voice devices with sound pickup angles in the range of 60 to 120 degrees is greater than 1;
  • the first calculation unit is used to: if the number of voice devices with a sound pickup angle in the range of 60 to 120 degrees is greater than 1, then based on the pick-up angle and pickup angle of each voice device with a sound pickup angle in the range of 60 to 120 degrees Based on the intensity of the sound, calculate the arousal score corresponding to the voice device; the arousal score of a voice device is used to represent the probability of the user waking up the voice device; the voice device with the highest arousal score is used as the responding voice device;
  • the first determination unit is configured to: if the number of voice devices with a sound pickup angle in the range of 60 to 120 degrees is 1, use the voice device with a sound pickup angle in the range of 60 to 120 degrees as the responding voice device; if If the number of voice devices with sound pickup angles in the range of 60 to 120 degrees is 0, then the voice device with the highest sound pickup intensity will be used as the responding voice device.
  • the first calculation unit is specifically configured to use a second calculation formula to calculate the arousal score corresponding to the voice device, and the second calculation formula includes:
  • P2 is the arousal score
  • s is the sound pickup intensity
  • r is the pickup angle
  • the first range is: r is greater than or equal to 60 and less than 80, or r is greater than 100 and less than or equal to 120
  • the second range is: r is greater than or equal to 80 and less than or equal to 100
  • a and b are preset weights, and a is greater than b, and c is 10*a+b/10.
  • the device provided in the second aspect corresponds to the method provided in the first aspect.
  • examples, beneficial effects, etc. of the relevant content in this aspect please refer to the relevant content in the first aspect, and will not be repeated here. Repeat.
  • embodiments of the present invention provide a wake-up system based on collaborative error correction.
  • Figure 8 which includes voice devices distributed in a home and a cloud platform that communicates with each voice device; the voice device includes Smart home appliances, the voice module in the smart home appliances has voice pickup and voice feedback functions; the cloud platform is used to control voice devices in the entire space; the decision-making device is pre-selected from the voice devices In a smart home appliance, the decision-making device has a wake-up device based on collaborative error correction provided in the second aspect.
  • the wake-up system includes each voice device, a cloud platform, and of course, an application program on a mobile terminal, which can control each voice device.
  • N in Figure 8 is a positive integer greater than 2.
  • embodiments of the present invention provide a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, the steps of the method provided in the first aspect are implemented.
  • a system or device equipped with a storage medium may be provided, on which the software program code that implements the functions of any of the above embodiments is stored, and the computer (or CPU or MPU) of the system or device ) reads and executes the program code stored in the storage medium.
  • the program code itself read from the storage medium can implement the functions of any one of the above embodiments, and therefore the program code and the storage medium storing the program code form part of the present invention.
  • Examples of storage media for providing program codes include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), Tape, non-volatile storage Memory card and ROM.
  • the program code can be downloaded from the server computer via the communications network.
  • the program code read from the storage medium is written into the memory provided in the expansion board inserted into the computer or written into the memory provided in the expansion module connected to the computer, and then based on the program code
  • the instructions cause the CPU installed on the expansion board or expansion module to perform part or all of the actual operations, thereby realizing the functions of any of the above embodiments.
  • embodiments of the present invention provide a voice device, including a memory, a processor, and a computer program stored in the memory and executable on the processor.
  • the processor executes the computer program, the first aspect is implemented. The steps of the method are provided.
  • the voice device is the decision-making device.
  • embodiments of the present invention also provide a complete set of home appliance control system.
  • the complete set of home appliance control system includes an Internet of Things platform 1, a content platform 2, a semantic platform 3, and a cloud platform 4.
  • voice acquisition module 7, smart terminal 5, the Internet of Things platform 1 can be the Zhihuijia AIOT platform.
  • the Internet of Things platform has a configuration module 11 and an intelligent control scene module 12.
  • the configuration module 11 and the intelligent control scene module 12 communicate through The network transmits information with the intelligent terminal 5 and controls the operation of the intelligent terminal 5; the cloud platform 4 transmits information with the semantic platform 3 and the Internet of Things platform 1 through the communication network respectively; the mobile terminal 6 implements information transmission with the cloud platform 1 through wireless means ;
  • the semantic platform 3 includes a speech analysis module 33, a functional feedback module 32 and an interactive feedback module 31; the semantic platform 3 can analyze the speech signal obtained by the speech acquisition module 7, and output control instructions according to the analyzed speech signal, and the control instructions are transmitted through wireless communication Communicate with cloud platforms and IoT platforms to further control or drive the normal operation of smart terminals.
  • this complete set of home appliance control system includes a mobile terminal 6 and a voice broadcast module 8.
  • the mobile terminal can communicate with the Internet of Things platform through the APP, or the mobile terminal 6 can communicate with the Zhihuijia AIOT platform Carry out communication and control the normal operation of the intelligent terminal; the voice broadcast module 8 completes the voice broadcast.
  • the voice command system of the above-mentioned complete home appliance system includes a voice pickup module 71, a voice command analysis module 34, a logical judgment module 9, a designated equipment module 91, a non-designated equipment module 92, an AI hierarchical management module 311, and an AI intelligent grouping module 312.
  • the voice broadcast module 81 the smart device command execution module 10 and the feedback command execution result module 11; when the voice pickup module obtains the user's voice command, it transmits the voice command to the voice command analysis module 33, and then passes the logic judgment module 9
  • the judgment result drives and controls the operation of the intelligent terminal, and the intelligent terminal feeds back and broadcasts the execution results of the voice instructions;
  • the AI hierarchical management module 311 can perform hierarchical management, and the AI intelligent grouping module 312 can implement intelligent grouping;
  • the designated device module 91 can perform specific The intelligent terminal performs designated management and control, and the non-designated device module 92 can perform unified management and control on the non-specific intelligent terminal.
  • the specified equipment can be defined by spatial location (such as room location) and equipment name (air conditioner or refrigerator, etc.).
  • the logical judgment module can control smart terminals according to priority;
  • the AI hierarchical management module can classify and manage the functions of air conditioners, refrigerators, washing machines and TVs according to the functions of smart terminals. This classification can be divided into general skills, exclusive skills and public skills.
  • the AI intelligent grouping module can manage and control intelligent terminals (such as air conditioners, refrigerators, washing machines and TVs) in the space area according to the space area (living room, bedroom, study room, kitchen, balcony).
  • the voice pickup module 71 and the voice acquisition module may be one module, or of course may not be the same module.
  • the voice command analysis module 34 and the voice analysis module 33 may be the same module, or of course may not be the same module.
  • the voice playback module 8 and the voice playback module 81 may be one module, or of course they may not be the same module. a module.
  • the functions described in the present invention can be implemented using hardware, software, plug-ins, or any combination thereof.
  • the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Human Computer Interaction (AREA)
  • Acoustics & Sound (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Automation & Control Theory (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Theoretical Computer Science (AREA)
  • Selective Calling Equipment (AREA)

Abstract

一种基于协同纠错的唤醒方法、装置及系统、介质、设备，该方法包括：在接收到一个语音设备发送来的唤醒请求时，判断该语音设备所在的空间分组内是否存在其它语音设备（S10）；若存在其它语音设备，则开始计时，在计时时长达到预设时长后判断在计时过程中是否接收到其它语音设备发送来的唤醒请求（S20）；若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送该唤醒请求的语音设备中确定一个应答语音设备，并向该应答语音设备发送唤醒应答指令，以使该应答语音设备从唤醒等待状态进入唤醒状态（S30）。该方法可以降低家庭内全部语音设备的整体误唤醒的概率。

Description

基于协同纠错的唤醒方法、装置及系统、介质、设备 技术领域
本说明书一个或多个实施例涉及语音设备技术领域，尤其是一种基于协同纠错的唤醒方法、装置及系统、介质、设备。
背景技术
语音识别技术是当下应用最广泛的人机交互技术，用户通过发起语音指令去控制语音设备执行相应的操作，比如，控制空调开机、关机。语音交互控制的全流程包括：语音唤醒、发送语音指令、设备执行动作、设备语音反馈等几个关键环节。语音唤醒即预先为语音设备设定一个或多个专有的词汇，用户通过该专有词汇可以将处于唤醒等待状态的语音设备激活，进入语音指令识别等待状态，用户可进一步发起各类语音指令，语音设备在收到后，执行对应的指令动作，并通过设备自带的播放部件，反馈指令执行的结果，提示用户。
当用户说出专有词汇时，语音设备可被激活，说出非专有词汇时，语音设备保持唤醒等待，避免用户正常的人际之间的语音交流误触发语音。但是在实际场景中仍然存在语音设备被非专有词汇、甚至被环境噪音唤醒激活的情况，行业内称该类唤醒为误唤醒，概率指标一般在1-3次/72小时不等。误触发会打扰到用户，拉低用户体验。
随着语音识别的广泛应用，越来越多家用电器装配了语音识别系统，各自独立的提供语音人机交互功能。当同一用户家庭中有多个语音设备时，特别是同一房间内有数个体语音设备时，考虑到单个语音设备的独立误唤醒概率，该房间内的语音设备整体误唤醒概率将成倍增加。
发明内容
本说明书一个或多个实施例描述了一种基于协同纠错的唤醒方法、装置及系统、介质、设备。
第一方面，本说明书提供了一种基于协同纠错的唤醒方法，在一个家庭的全空间内分布有语音设备，所述语音设备中包括智能家电，所述智能家电中的语音模块具有语音拾取和语音反馈功能；各个所述语音设备均与云平台通信连接；所述方法由决策设备执行，所述决策设备为预先从所述语音设备中选出的一个智能家电；所述方法包括：
在接收到一个语音设备发送来的唤醒请求时，判断该语音设备所在的空间分组内是否存在其它语音设备；其中，一个处于唤醒等待状态的语音设备监听到唤醒词后向所述决策设备发送唤醒请求，一个家庭的全空间内的各个语音设备具有相同的唤醒词；
若存在其它语音设备，则开始计时，在计时时长达到预设时长后判断在计时过程中是否接收到所述其它语音设备发送来的唤醒请求；
若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送所述唤醒请求的语音设备中确定一个应答语音设备，并向所述应答语音设备发送唤醒应答指令，以使所述应答语音设备从唤醒等待状态进入唤醒状态。
第二方面，本说明书提供了一种基于协同纠错的唤醒装置，在一个家庭的全空间内分布有语音设备，所述语音设备中包括智能家电，所述智能家电中的语音模块具有语音拾取和语音反馈功能；各个所述语音设备均与云平台通信连接；
所述装置安装在决策设备上，所述决策设备为预先从所述语音设备中选出的一个智能家电；所述装置包括：
第一判断模块，用于在接收到一个语音设备发送来的唤醒请求时，判断该语音设备所在的空间分组内是否存在其它语音设备；其中，一个处于唤醒等待状态的语音设备监听到唤醒词后向所述决策设备发送唤醒请求，一个家庭的全空间内的各个语音设备具有相同的唤醒词；
第二判断模块，用于若存在其它语音设备，则开始计时，在计时时长达到预设时长后判断在计时过程中是否接收到其它语音设备发送来的唤醒请求；
目标确定模块，用于若在计时过程中接收到其它语音设备发送来的唤醒请求，从发 送所述唤醒请求的语音设备中确定一个应答语音设备，并向所述应答语音设备发送唤醒应答指令，以使所述应答语音设备从唤醒等待状态进入唤醒状态。
第三方面，本发明实施例提供一种基于协同纠错的唤醒系统，包括在一个家庭内分布设置的语音设备以及与各个语音设备通信连接的云平台；所述语音设备中包括智能家电，所述智能家电中的语音模块具有语音拾取和语音反馈功能；所述云平台用于对全空间内的语音设备进行控制；所述决策设备为预先从所述语音设备中选出的一个智能家电，所述决策设备具有第二方面提供的基于协同纠错的唤醒装置。
第四方面，本发明实施例提供一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现第一方面提供的方法的步骤。
第五方面，本发明实施例提供一种语音设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如第一方面提供的所述方法的步骤。
第六方面，本发明实施例提供一种成套家电控制系统，包括物联网平台、内容平台、语义平台、云平台、语音获取模块、智能终端；其中，物联网平台为智汇家AIOT平台，该物联网平台具有配置模块和智能控制场景模块，配置模块和智能控制场景模块通过通信网络与智能终端进行信息传输并控制智能终端的运行；云平台分别与语义平台和物联网平台通过通信网络进行信息传输；语义平台包括语音解析模块、功能反馈模块和交互反馈模块，语义平台能够解析语音获取模块获取的语音信号，并根据解析的语音信号输出控制指令，该控制指令通过无线通信的方式分别与云平台和物联网平台进行通信，控制或驱动智能终端的运行。
本说明书实施例提供的基于协同纠错的唤醒方法、装置及系统、介质、设备，具有以下有益效果：在本发明实施例中，在决策设备接收到一个语音设备发送来的唤醒请求时，决策设备会判断发送唤醒请求的语音设备所在的空间分组内是否还存在其它的语音设备，如果存在其它的语音设备，则开始计时，在计时结束后判断在计时过程中是否接收到其它语音设备的唤醒请求，如果接收到其它语音设备的唤醒请求，则确定应答语音设备，进而向应答语音设备发送唤醒应答指令，这样应答语音设备就会从唤醒等待状态进入到唤醒状态。这个过程用户是感知不到的，不会影响语音设备原本的工作模式，而且在空间分组内有多个语音设备时还通过本发明实施例提供的方法，将单一语音设备被误触发的异常状态排除掉，降低家庭内全部语音设备的整体误唤醒的概率。针对一次唤醒，决策设备接收到多个语音设备发送来的唤醒请求，此时决策设备会选出一个最合适的语音设备作为应答语音设备，可以为用户带来优质的使用体验。
附图说明
为了更清楚地说明本说明书实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图是本说明书的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是本说明书一个实施例中基于协同纠错的唤醒方法的流程示意图；
图2是本说明书一个实施例中四个语音设备接入同一个网络的示意图；
图3是本说明书一个实施例中四个语音设备的房间分配的示意图；
图4是本说明书一个实施例中语音唤醒处理方法的流程示意图；
图5是本说明书一个实施例中一个家庭中多个智能家电的分布示意图；
图6a是本说明书一个实施例中一个语音终端的示意图；
图6b是本说明书一个实施例中一个家庭内多个语音终端的分布示意图；
图7是本说明书一个实施例中基于协同纠错的唤醒装置的结构框图；
图8是本说明书一个实施例中基于协同纠错的唤醒系统的结构框图。
图9是本说明书一个实施例中成套家电系统的架构图；
图10是本说明书一个实施例中语音指令系统的控制原理示意图。
具体实施方式
下面结合附图，对本说明书提供的方案进行描述。
第一方面，本发明实施例提供一种基于协同纠错的唤醒方法。
本发明实施例提供的方案的适用场景可以包括但不限于如下场景：
在一个家庭内分布有多个语音设备，这些语音设备包括各种智能家电，例如，参见图5，在一个家庭的全空间内设置有一个洗衣机、一个冰箱、两个挂式空调、一个柜式空调、两个电视机。这些智能家电均具有语音模块，智能家电中的语音模块能够拾取用户发出的语音指令，把语音指令发送给云平台。语音模块也可以在云平台的控制下进行语音反馈，即通过语音播报的方式告知用户一些设备的执行情况。
但是由于在阳台、入户区域、卫生间等没有设置智能家电，因此这些区域的语音拾取功能会较差，因此可以在这些区域设置语音终端。参见图6a和6b，在入户区域、阳台、餐厅、书房、次卫、主卫、主卧的床头、儿童房的床头等位置均设置了语音终端。语音终端可以拾取用户发出的语音，也可以通过WIFI接入家庭局域网。由于语音终端的体积非常小，不便安装扩音模块，因此语音终端不具有语音播报的功能。
可理解的是，由于各个智能家电的最佳拾音范围很难将一个家庭内的各个角落都覆盖到，因此有时会出现不能对用户的语音指令进行响应的情况，为此本发明实施例在家庭内智能家电覆盖不到的区域内设置了语音终端。这样各个智能家电和各个语音终端的最佳拾音范围可以覆盖一个家庭内的全空间。
当然，语音终端也可以具有灯光提示的功能。例如，当一个语音终端被唤醒作为应答语音设备时可以通过呼吸灯的方式进行提示，当一个语音终端作为应答语音设备，云平台控制对应的目标智能家电执行指令成功后可以控制作为应答语音设备的语音终端采用绿灯快闪三次的方式进行提示，当云平台控制对应的目标智能家电执行指令失败时可以控制作为应答语音设备的语音终端采用黄灯快闪三次的方式进行提示。
其中，一个家庭内的所有语音设备(包括各个语音终端、各个智能家电)均与云平台通信连接，这样云平台可以对家庭内的各个语音设备进行控制。当然，用户的移动终端上可以安装一个应用程序，通过该应用程序也可以对家庭内各个语音终端的控制、配置等。
举例1，用户在阳台上发出唤醒词“长虹小白”，阳台上的语音终端、客厅的电视机、客厅的柜式空调都会拾取到唤醒词，那到底唤醒哪一个语音设备呢，这是本发明实施例要解决的问题。如果确定客厅的电视机被唤醒，客厅的电视机会发出回应“在的”，进而用户会发出语音指令“打开客厅空调”，此时客厅电视机会拾取该段语音，并将该段语音发送至云平台，云平台对该段语音进行解析，进而控制客厅的柜式空调打开。
可见，本发明实施例中被唤醒的语音设备被称为应答语音设备，该应答语音设备的作用是拾取用户发出的语音指令，将语音指令发送给云平台，这样云平台会对语音指令进行解析，根据解析结果控制执行设备进行相应的操作。
综上可知，在一个场景中，在一个家庭的全空间内分布有语音设备，所述语音设备包括智能家电和语音终端，各个所述智能家电和各个所述语音终端的最佳拾音范围的并集能够覆盖所述全空间；所述智能家电中的语音模块具有语音拾取和语音反馈的功能，所述语音终端的语音模块具有语音拾取的功能；所述语音设备均与用于进行全空间内语音设备控制的云平台通信连接。
本发明实施例提供了一种基于协同纠错的唤醒方法，该方法可以由决策设备执行，所述决策设备为预先从所述语音设备中选出的一个智能家电。该方法适用的场景除了上述场景之外，还可以是在一个家庭内仅包括智能家电的场景，该智能家电中的语音模块具有语音拾取和语音反馈功能。
可理解的是，如果本发明的一些可选实施例中涉及到了语音终端，则这样的具体实 施例适用的场景为：在一个家庭的全空间内分布有语音设备，所述语音设备包括智能家电和语音终端，各个所述智能家电和各个所述语音终端的最佳拾音范围的并集能够覆盖所述全空间；所述智能家电中的语音模块具有语音拾取和语音反馈功能，所述语音终端的语音模块具有语音拾取功能；所述语音设备均与用于进行全空间内语音设备控制的云平台通信连接。例如，在一个家庭内包含至少两个智能家电和至少两个语音终端。
可理解的是，如果本发明的一些可选实施例中没有涉及到语音终端，则这样的具体实施例适用的场景可以为上一段文字所描述的场景，当然也可以为场景：在一个家庭内仅包含智能家电的场景，该智能家电中的语音模块具有语音拾取和语音反馈功能。
参见图1，所述方法包括如下步骤S10～S30：
S10、在接收到一个语音设备发送来的唤醒请求时，判断该语音设备所在的空间分组内是否存在其它语音设备；
其中，一个处于唤醒等待状态的语音设备监听到唤醒词后向所述决策设备发送唤醒请求，一个家庭的全空间内的各个语音设备具有相同的唤醒词。
可理解的是，本发明实施例提供的方法由决策设备执行，而决策设备是一个家庭内的各个语音设备中的一个智能家电，因此在执行本发明实施例提供的方法之前，需要先确定出执行设备。
在一个家庭内会有多个语音设备，例如，智能语音冰箱、语音音箱、语音空调、语音电视、多个语音终端等。在一个家庭内的各个语音设备的唤醒词都是一样的，也就是说，一个家庭内的各个语音设备都是基于同一个唤醒词进行训练的。而且，在移动终端上可以通过一个应用程序对这个家庭内的各个语音设备进行控制，该应用程序就是下文中提到的预设应用程序。决策设备的确定可以由这个预设应用程序实现。
也就是说，各个语音设备均由所在家庭的用户移动终端上的同一个预设应用程序控制；所述决策设备由所述预设应用程序预先确定；其中，所述预设应用程序确定所述决策设备的过程包括如下步骤S01～S04：
S01、获取接入家庭网络的各个语音设备；
可理解的是，当每一个语音设备接入家庭网络时，在预设应用程序上就会看到这个语音设备的相关信息，例如，设备标识、设备类型、设备名称等。因此预设应用程序可以获得各个接入家庭网络的语音设备的相关信息所形成的设备列表。
S02、获取加入所述家庭网络的每一个语音设备的带电状态数据、运算能力数据和用户使用习惯数据；
可理解的是，在语音设备运行一段时间后，会产生很多数据，例如，用户对语音设备的使用习惯数据、语音设备的带电状态数据，这些数据会记录在语音设备内部的存储器中。预设应用程序会从每一个语音设备中获取到这些数据。同时，预设应用程序还会获取每一个语音设备的运算能力数据。
其中，用户对语音设备的使用习惯数据，例如，用户习惯在一天的哪个时间段看电视、用户习惯在哪个季节打开空调等。语音设备的带电状态数据是指语音设备是否一直带电，还是在一天中的某个时间段带电，还是很少带电等。语音设备的运算能力数据是指能够体现语音设备的运算速度、运算量的数据，因为决策设备在实现自身功能的同时还需要进行唤醒的决策处理，因此这里需要考虑到语音设备的运算能力。
S03、根据每一个语音设备的带电状态数据、运算能力数据和用户使用习惯数据，确定每一个语音设备的决策能力评分；
可理解的是，在计算每一个语音设备的决策能力评分时，不仅需要考虑用户的使用习惯、带电状态，还考虑语音设备的运算能力，即不仅考虑语音设备的外部因素，也考虑语音设备的内部因素，最后可以计算出一个体现语音设备的综合决策能力的评分。
在具体实施时，S03中具体可以采用第一计算式计算每一个语音设备的决策能力评分，所述第一计算式包括：
P1＝u*(d2-1/y)
式中，P1为所述决策能力评分；d为语音设备在过去一个月内的平均每天带电时长；y为语音设备的CPU运算能力；u为家庭用户在当前季节对语音设备的使用标记；若家庭用户在当前季节使用该语音设备，则该语音设备对应的使用标记为1；若家庭用户在当前季节不适用该语音设备，则该语音设备对应的使用标记为0。
在上述第一计算式中，u为用户的使用习惯数据。如果家庭用户在当前季节不使用该语音设备，则该语音设备对应的使用标记为0，此时P1为0。而如果家庭用户在当前季节中使用该语音设备，则该语音设备对应的使用标记为1，此时P1＝d2-1/y。例如，针对语音空调这一语音设备，有的家庭在冬天时不使用，只有在夏天的时候才使用。
其中，d为语音设备在过去一个月内的平均每天带电时长，该参数既考虑了用户在最近一段时间内的使用情况，也反映了在每一天的平均带电情况。例如，用户在过去一个月内只在晚上才打开语音电视，这样语音电视在过去一个月内的平均带电时长只有几个小时。而语音冰箱一直处于带电状态，此时语音冰箱的在过去一个月内的平均每天带电时长为24小时。参数d是一个关键的参数，语音设备的带电时长越长，语音设备能够进行唤醒决策处理的时间越多，可以减少遗漏唤醒请求的情况发生。d越大，P1越大。
其中，y体现的是语音设备的运算能力，例如，cpu的运算能力。不同的cpu的运算能力的计算方式不同，例如，有的cpu通过字长衡量运算能力，有的cpu通过双精度浮点运算能力来衡量运算能力，具体可以根据cpu的实际情况而确定。y越大，P1越大，但y和P1之间并不是正比的关系。
可见，上述第一计算式可以非常合理的反映出一个语音设备的决策能力。
S04、将所述决策能力评分最高的语音设备作为所述决策设备。
也就是说，在S03中，计算出每一个接入家庭网络的语音设备的决策能力评分，然后选择出决策能力评分最高的语音设备作为决策设备。
在具体实施时，在通过上述方式确定出一个决策设备之后，可以采用该决策设备执行本发明实施例提供的方法。但是如果决策设备掉电，则需要更换其它的语音设备作为决策设备。当然，也可能存在之前确定出的决策设备不带电的情况，此时也需要更换其它的语音设备作为决策设备。
由于在实际场景中可能发生以上情况时，因此所述预设应用程序确定所述决策设备的过程还可以包括：所述预设应用程序通过心跳指令的方式按照预设时间间隔检测当前的决策设备是否处于带电状态；若未处于带电状态，则选择处于带电状态的各个语音设备中决策能力评分最高的一个语音设备作为当前的决策设备。
也就是说，移动终端上的预设应用程序会按照时间间隔检测当前的决策设备是否带电，具体通过心跳指令的方式检测，例如，移动终端上的预设应用程序会每间隔一段时间就向决策设备发送一个心跳指令，如果移动终端能够接收到决策设备的反馈信息，则说明此时决策设备此时带电，如果不能接收到决策设备的反馈信息，说明此时决策设备不带电。
如果移动终端上的预设应用程序经过检测发现当前的决策设备不带电，需要更换其它的语音设备，此时会在各个带电的语音设备中选择出决策能力评分最高的语音设备作为当前的决策设备，这样可以避免因为决策设备不带电而影响唤醒处理。
当然，移动终端上的预设应用程序除了可以自动确定决策设备之外，用户也可以在预设应用程序上对每一个语音设备所在的房间进行标记，即为每一个语音设备设置对应的房间标识。
在具体实施时，上述空间分组可以为物理空间分组，也可以为虚拟空间分组。当上述空间分组为物理空间分组时，一个房间对应一个空间分组，而一个空间分组内包括至少一个语音设备。也就是说，所述空间分组为按照物理空间分组的方式对各个房间内的语音设备进行划分后得到的多个组。
具体的，可以在预设应用程序上对各个设备进行分组。具体的，所述预设应用程序可以用于提供配置界面，以使用户对接入家庭网络的每一个语音设备配置对应的房间标识；对应的，S10中所述判断该语音设备所在的空间分组内是否存在其它语音设备，包括：从所述预设应用程序中获取该语音设备对应的房间标识，将该房间标识作为第一房间标识；判断其它语音设备的房间标识中是否存在与所述第一房间标识相同的房间标识；若存在，则该语音设备所在的房间内存在其它语音设备；否则，该语音设备所在的房间内不存在其它语音设备。
可理解的是，用户可以在预设应用程序的配置界面上对每一个语音设备设置对应的房间标识，这样预设应用程序就可以知道每一个房间内都有哪些语音设备。进一步的，当决策设备判断该语音设备所在的空间分组内是否存在其它语音设备时，会从移动终端的预设应用程序上获取发送唤醒请求的这个语音设备对应的房间标识即第一房间标识，进而判断其它的语音设备的房间标识中是否存在与第一房间标识相同的房间标识，进而判断出在发送唤醒请求的语音设备所在的房间中是否还存在其它的语音设备。
在实际场景中，如果用户发出了唤醒词，在一个房间内的处于唤醒等待状态的语音设备监听到这个唤醒词之后，会向决策设备发送唤醒请求，进而决策设备判断该语音设备所在的房间内是否还存在其它的语音设备，进而执行后续的步骤。
可理解的是，如果发出唤醒请求的语音设备是决策设备本身，处理方法也是类似的。
S20、若存在其它语音设备，则开始计时，在计时时长达到预设时长后判断在计时过程中是否接收到其它语音设备发送来的唤醒请求；
可理解的是，由于各个语音设备的唤醒词都是一样的，而如果在一个房间内除了发出唤醒请求的预设设备之外，还存在其它语音设备，此时决策设备就会开始计时，当计时时长达到预设时长后停止计时，在计时的过程中有可能会接收到其它语音设备发送来的唤醒请求，也有可能没有接收到其它语音设备发送来的唤醒请求。
S30、若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送所述唤醒请求的语音设备中确定一个应答语音设备，并向所述应答语音设备发送唤醒应答指令，以使所述应答语音设备从唤醒等待状态进入唤醒状态；
进一步的，S30中所述若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送所述唤醒请求的语音设备中确定一个应答语音设备，包括：
若在计时过程中接收到其它语音设备发送来的唤醒请求，则获取每一个发送唤醒请求的语音设备的拾音参数；
根据发送唤醒请求的各个语音设备各自的拾音参数，从发送所述唤醒请求的语音设备中确定所述应答语音设备。
可理解的是，如果接收到其它语音设备发送来的唤醒请求时，说明不止一个语音设备针对用户发生的语音做出了反映，此时是误唤醒的可能性会比较小，因此此时需要在各个发出唤醒请求的语音设备中选择一个概率最高的唤醒对象。而如果在计时过程中没有接收到其它语音设备的唤醒请求，则发出唤醒请求的语音设备被误唤醒的可能性是很大。
如果在计时过程中，接收到了其它语音设备发送来的唤醒请求，此时可以获取每一个发送唤醒请求的拾音参数，进而利用拾音参数进行下一步的计算。所谓的拾音参数是指语音设备所采集用户发出的语音时的一些参数。
其中，所述拾音参数可以包括拾音角度和拾音声强，如果用户会对着想要唤醒的语音设备发出语音，此时的拾音角度为90°，但是用户没有对着想要唤醒的语音设备发出语音，此时的拾音角度不是90°，可见拾音角度反映的是用户发出语音时语音设备的声音拾取角度。拾音声强是指语音设备所拾取的声音的强度。
基于上述拾音参数，根据发送唤醒请求的各个语音设备各自的拾音参数，确定应答语音设备，具体过程可以包括如下步骤S41～S43：
S41、针对发送唤醒请求的各个语音设备，判断拾音角度在60～120度范围内的语音设备的数量是否大于1；
其中，60～120度是一个较佳的拾音角度，该拾音角度在60～120度时语音设备是应答语音设备的可能性是很大的。
可理解的是，在该步骤中首先确定出拾音角度在60～120度范围内的语音设备的数量，如果该数量大于1，则需要在这些拾音角度在60～120度范围内的各个语音设备中进一步筛选。
S42、若拾音角度在60～120度范围内的语音设备的数量大于1，则根据每一个拾音角度在60～120度范围内的语音设备的拾音角度和拾音声强，计算该语音设备对应的唤醒评分；一个语音设备的唤醒评分用于表征用户唤醒该语音设备的概率；
在该步骤中，针对拾音角度在60～120度范围内的各个语音设备，根据语音设备的拾音角度和拾音声强，计算每一个语音设备的唤醒评分，从而得知用户想要唤醒该语音设备的概率是多少。
进一步的，该步骤S42中可以采用第二计算式计算该语音设备对应的唤醒评分，所述第二计算式包括：
当r在第一范围内时，P2＝a*s+b/|r-90|
当r在第二范围内时，P2＝b*s-a*|r-90|+c
式中，P2为所述唤醒评分，s为所述拾音声强，r为所述拾音角度，所述第一范围为：r大于等于60且小于80，或者，r大于100且小于等于120；所述第二范围为：r大于等于80且小于等于100；a和b为预设权重，且a大于b，c为10*a+b/10。
其中，当r在第一范围内时，拾音强度越大，P2越大；拾音角度越接近90，P2越大，所以拾音强度高且拾音角度接近90的语音设备的唤醒评分是比较高的。而且，当r在第一范围内时，拾音声强的权重为a，而a大于b，说明此时更为关注拾音声强，也就是说，在这种情况下拾音声强更为重要。
其中，当r在第二范围内时，拾音强度越大，P2越大，拾音角度越接近，P2越大，所以拾音强度高且拾音角度接近90的语音设备的唤醒评分是比较高的。当r在第二范围内时，拾音角度的权重为a，而a大于b，说明此时更为关注拾音角度，也就是说，在这种情况下，拾音角度跟为重要。
进一步的，当拾音声强相同，而r在第二范围时的唤醒评分应大于r在第一范围内的唤醒评分，为保证此要求，本发明实施例中当r在第二范围内时的计算式中增加了参数c，且令c为10*a+b/10，该值可以保证在拾音声强相同的情况下r在第二范围时的唤醒评分大于r在第一范围内的唤醒评分。
其中，c＝10*a+b/10是当r为100时、拾音声强为0时，为保证b*s-a*|r-90|+c大于等于a*s+b/|r-90|而计算得到。在该c值的基础上，当r处于第二范围内的任一值时，都能保证在拾音声强相同的情况下r在第二范围时的唤醒评分大于r在第一范围内的唤醒评分。
S43、将唤醒评分最高的语音设备作为应答语音设备。
可理解的是，一个语音设备的唤醒评分越高，说明用户想要唤醒这个语音设备的概率越大，如果一个语音设备的唤醒评分越低，说明用户想要唤醒这个语音设备的概率越低。因此这里将唤醒评分最高的语音设备作为应答语音设备。
在具体实施时，根据发送唤醒请求的各个语音设备各自的拾音参数，确定应答语音设备，具体过程还可以包括如下至少一项：
(1)若所述拾音角度在60～120度范围的语音设备的数量为1，则将所述拾音角度在60～120度范围的语音设备作为应答语音设备；
可理解的是，如果拾音角度在60～120度范围内的语音设备的数量只有一个，则这一个语音设备是用户想要唤醒的语音设备的概率是最大的，在这种情况下不需要考虑语音设备的拾音声强，因此此时将这一个语音设备作为应答语音设备。
(2)若所述拾音角度在60～120度范围的语音设备的数量为0，则将所述拾音声强最高的语音设备作为应答语音设备。
可理解的是，如果拾音角度在60～120度范围的语音设备的数量为0，说明用户没有对着任何一个语音设备发出语音，此时不需要考虑拾音角度，此时可以在发出唤醒请求的各个语音设备中选择拾音声强最大的语音设备作为应答语音设备。
可见，在不同的情况下可以依照上述过程确定应答语音设备。
可理解的是，以上均是在计时过程中还接收到其它语音设备发送来的唤醒请求时的处理方法。在实际中如果在计时过程中没有接收到其它语音设备发送来的唤醒请求的话，可以认为发送唤醒请求的语音设备为误触发，此时决策设备可以向发出唤醒请求的语音设备发送终止唤醒应答指令，接收到终止唤醒应答指令的语音设备不会从唤醒等待状态进入唤醒状态，而是继续保持唤醒等待状态。
可理解的是，以上均是针对在发出唤醒请求的语音设备所在的空间分组还存在其它语音设备的情况。而当发出唤醒请求的语音设备所在空间分组内没有其它语音设备时，本发明实施例提供的方法还可以包括如下步骤：向发送唤醒请求的语音设备发送唤醒应答指令，以使语音设备进入唤醒状态。
也就是说，在一个空间分组内只有一个语音设备，而该语音设备向决策设备发出了唤醒请求，此时决策设备认为用户就是想要唤醒这个语音设备，因此决策设备会向该设备发出唤醒应答指令，接收到该指令的语音设备会从唤醒等待状态进入唤醒状态。
可理解的是，本发明实施例中的决策设备可以和发出唤醒请求的语音设备在一个房间内，也可以不和发出唤醒请求的语音设备在一个空间分组内。
可理解的是，本发明实施例中的各个语音设备接入同一个网络中，例如，通过以太网、蓝牙、WIFI等方式接入同一个家庭网络中。这样各个语音设备可以相互进行通信，传递数据，
举例来说，参见图2，在一个家庭内有语音设备A、语音设备B、语音设备C、语音设备D，语音设备A、语音设备B采用WIFI网络技术，语音设备C采用有线以太网技术，语音设备D采用蓝牙通信技术，均接入家庭网络设备中，并选定了语音设备A为决策设备。
用户在其手机的应用程序中标记了每一个语音设备的房间标识，例如，参见图3，通过APP软件将语音设备A、语音设备B、语音设备C分配在房间A中，将语音设备D分配在房间B中，将房间A和房间B的设备列表及设备ID信息下发至语音设备A。
参见图4，当语音设备B被用户的语音触发，需要唤醒时，首先通过语音设备B网络向决策设备即语音设备A发起唤醒请求，决策设备检查到语音设备B的房间A分组内还有语音设备A和语音设备C，则进入计时等待，计时结束后判断语音设备A和语音设备C是否也发起了唤醒请求。经过判断本次只有语音设备B提交了唤醒请求，所以本次判断为误唤醒，决策设备即向语音设备B发出终止应答指令。
以上流程判断过程中，用户均不会感知，有效降低因多语音设备的误唤醒带来的影响。
可见，本发明实施例提供的方法是针对在一个家庭内存在多个语音设备的情况下误触发的概率成倍增加的问题，提出的一种支持协同纠错的语音唤醒方案，在该方案中需要先进行设备入网、选出决策设备、标记房间，进而再执行本发明实施例提供的上述步骤S1～S4。
在具体实施时，空间分组还可以是虚拟空间分组。所述空间分组为按照发出唤醒词的用户在所述全空间内的位置和朝向进行划分后得到的虚拟空间分组，一个虚拟空间分组内包括至少两个语音设备；所述空间分组为所述云平台预先确定。虚拟空间分组并不是有用户在应用程序上自行设置的，而是云平台依据用户的位置和朝向划分的，不需要用户设置。
针对虚拟空间分组，云平台确定空间分组的方法包括：
a1、在历史唤醒任务中获取发出唤醒词的用户在所述全空间中的位置和朝向；
a2、获取各个语音设备拾取到所述预设唤醒词的声音强度，并根据所述声音强度从各个语音设备中选出第一语音设备；其中，各个所述第一语音设备所拾取到的声音强度高于其它的语音设备所拾取到的声音强度，且各个所述第一语音设备所拾取到的声音强度之间的最大差值在预设范围内；
a3、获取各个所述第一语音设备的拾音角度，并根据各个所述第一语音设备的拾音角度，从所述第一语音设备中选出拾音角度落在最佳拾音角度范围内的第二语音设备；
a4、将各个所述第二语音设备形成针对用户的所述位置和所述朝向的空间分组。
举例来说，用户在客厅和餐厅的中间位置并面向阳台发出“长虹小白”，此时入户、客厅和餐厅内的5个语音设备拾取到的声音强度稍微高于洗衣房和厨房内的语音设备拾取到的声音强度，入户、客厅和餐厅内的5个语音设备拾取到的声音强度远高于在书房、卧室和卫生间中的语音设备拾取到的声音强度。根据各个语音设备拾取到的声音强度筛选出第一语音设备。由于用户站在客厅和餐厅的中间位置，此时入户、客厅和餐厅内的5个语音设备拾取的声音强度是差不多的，而且这5个语音设备是所有语音设备中声音强度最大的语音设备，因此将这5个语音设备作为第一语音设备。
进一步的，由于用户面向阳台，因此位于客厅和阳台中的3个语音设备面向用户的发声方向，而入户、餐厅内的两个语音设备背向用户的发声方向，因此客厅和阳台中的3个语音设备的拾音角度位于最佳拾音角度范围，而入户、餐厅内的两个语音设备的拾音角度未处于最佳拾音角度范围内，因此将客厅和阳台中的3个语音设备作为第二语音设备，这3个第二语音设备形成一个虚拟空间分组。当然，如果用户面向的是餐厅方向，则餐厅和入户的两个语音设备形成一个虚拟空间分组。
可见，依据用户的位置和朝向可以对应一个虚拟空间分组，实际上当用户在某一个位置时就可以对应一个虚拟空间分组，例如，用户处于坐在客厅沙发上看电视时，此时对应的虚拟空间分组为客厅和阳台的三个语音设备形成的虚拟空间分组。
在实际中，用户处于家庭的全空间内的不同位置和朝向时，针对多次的唤醒请求，可以形成多个虚拟空间分组，将多个虚拟空间分组进行保存，在后续可以直接使用虚拟空间分组。
本发明实施例提出了一个虚拟空间分组。虚拟空间分组是根据以往的经验自动划分得到的虚拟空间分组。虚拟空间分组是依据用户的位置和朝向所确定的设备分组，一个虚拟空间分组内的各个语音设备可以位于同一个物理空间内，也可以为位于多个物理空间内。例如，当用户站在客厅和餐厅的中间并朝向阳台的方向发出唤醒请求时，在客厅的柜式空调和电视机、位于阳台的语音终端这三个设备会被划分到一个虚拟空间分组中。虚拟空间分组为一个虚拟空间分组，在这个虚拟空间分组中有至少两个语音设备，如果仅有一个语音设备就没有必要划分为一个组了。
可理解的是，在决策设备接收到一个语音设备发送来的唤醒请求时，判断在该语音设备所在的虚拟空间分组内是否有其它的语音设备。如果存在其它的语音设备，则判断在一定的时间内是否能够接收到其它语音设备发送来的唤醒请求。如果接收到其它语音设备发送来的唤醒请求，说明此时为误唤醒的可能性比较小，进而在发送唤醒请求的各个语音设备中选择一个语音设备作为应答语音设备。但是如果在所述虚拟空间分组内存在其它的语音设备，但是在一定的时间段内却没有接收到其它语音设备发送来的唤醒请求，说明误唤醒的可能性比较大，此时不确定应答语音设备。
当然，由于一个虚拟空间分组内存在至少两个语音设备，因此当一个虚拟空间分组内的一个语音设备发送给决策设备一个唤醒请求后，不存在该虚拟空间分组内不存在其它语音设备的情况。只有在物理空间分组时，才会出现一个物理分组空间内只有一个语音设备的情况。
针对虚拟空间分组，有可能出现一个语音设备不具有对应的虚拟空间分组的情况，例如，在一个房间内只有一个语音设备。在云平台确定虚拟空间分组时，没有将该语音设备划分到任何一个虚拟空间分组中。针对一个语音设备不具有对应的虚拟空间分组的情况，当决策设备接收到这个语音设备发送来的唤醒请求时，会直接将该语音设备作为应答语音设备。
可理解的是，在一个家庭的全空间内为了能够拾取到用户在各个角落发出的语音指令，设置了多个语音终端，使智能家电和语音终端的最佳拾音范围的并集能够覆盖全空间，这样可以避免或大大减少因没有拾取到用户指令而没有对用户指令进行响应的问题。而且各个语音终端与云平台连接，这样可以形成针对一个家庭的全空间内的各个语音终端的控制。
在一个实施例中，本发明实施例提供的方法可以是一种协同纠错的就近唤醒方法，该就近唤醒方法的步骤和上述步骤S10～S30一致，进一步的，在确定应答语音设备时，可以具体是基于就近原则从发送所述唤醒请求的语音设备中确定一个应答语音设备。所谓的就近原则，例如，从发送唤醒请求的语音设备中选择距离用户最近的一个语音设备作为应答语音设备。再例如，在符合声音强度要求和拾音角度范围要求的多个语音设备中选择一个距离用户最近的语音设备作为应答语音设备。
在一个实施例中，当用户发出“长虹小白”或“海尔小黑”等唤醒词时，也可以直接唤醒决策设备，由决策设备进行去唤醒和控制其他的语音设备。
当然，在一个家庭内的各个语音设备也可以具有不同的唤醒词，但是上述方法是针对具有相同唤醒词的各个语音设备而执行的。
上文中，语音终端不具有语音反馈功能。当如果在语音终端中集成了喇叭模块，使得语音终端具有语音反馈功能的话，此时语音终端可以看作是一个智能家电，因为一般语音终端都具有提示灯的功能，此时语音终端可以看作是一个具有提示灯功能的智能家电。
上文中，语音设备可以直接与云平台通信连接，也可以通过家庭局域网与云平台进行通信连接。
当然，在上文中，拾音角度范围60度～120度，也可以扩展为50～130、40～140、30～150、20～160、10～170等。
本发明实施例提供的方法，在本发明实施例中，在决策设备接收到一个语音设备发送来的唤醒请求时，决策设备会判断发送唤醒请求的语音设备所在的空间分组内是否还存在其它的语音设备，如果存在其它的语音设备，则开始计时，在计时结束后判断在计时过程中是否接收到其它语音设备的唤醒请求，如果接收到其它语音设备的唤醒请求，则确定一个应答语音设备，进而向应答语音设备发送唤醒应答指令，这样应答语音设备就会从唤醒等待状态进入到唤醒状态。这个过程用户是感知不到的，不会影响语音设备原本的工作模式，而且在空间分组内有多个语音设备时还通过本发明实施例提供的方法，将单一语音设备被误触发的异常状态排除掉，降低家庭内全部语音设备的整体误唤醒的概率。针对一次唤醒，决策设备接收到多个语音设备发送来的唤醒请求，此时决策设备会依据拾音参数选出一个最合适的语音设备作为应答语音设备，在实现就近唤醒的同时，可以为用户带来优质的使用体验。
第二方面，本发明实施例提供一种基于协同纠错的唤醒装置，在一个家庭的全空间内分布有语音设备，所述语音设备中包括智能家电，所述智能家电中的语音模块具有语音拾取和语音反馈功能；各个所述语音设备均与云平台通信连接；所述装置安装在决策设备上，所述决策设备为预先从所述语音设备中选出的一个智能家电；参见图7，所述装置包括：
第一判断模块，用于在接收到一个语音设备发送来的唤醒请求时，判断该语音设备所在的空间分组内是否存在其它语音设备；其中，一个处于唤醒等待状态的语音设备监 听到唤醒词后向所述决策设备发送唤醒请求，一个家庭的全空间内的各个语音设备具有相同的唤醒词；
第二判断模块，用于若存在其它语音设备，则开始计时，在计时时长达到预设时长后判断在计时过程中是否接收到其它语音设备发送来的唤醒请求；
目标确定模块，用于若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送所述唤醒请求的语音设备中确定一个应答语音设备，并向所述应答语音设备发送唤醒应答指令，以使所述应答语音设备从唤醒等待状态进入唤醒状态。
在一个实施例中，所述拾音参数包括拾音角度和拾音声强；所述目标确定模块具体包括：
第一判断单元，用于：针对发送唤醒请求的各个语音设备，判断拾音角度在60～120度范围内的语音设备的数量是否大于1；
第一计算单元，用于：若拾音角度在60～120度范围内的语音设备的数量大于1，则根据每一个拾音角度在60～120度范围内的语音设备的拾音角度和拾音声强，计算该语音设备对应的唤醒评分；一个语音设备的唤醒评分用于表征用户唤醒该语音设备的概率；将唤醒评分最高的语音设备作为应答语音设备；
第一确定单元，用于：若所述拾音角度在60～120度范围的语音设备的数量为1，则将所述拾音角度在60～120度范围的语音设备作为应答语音设备；若所述拾音角度在60～120度范围的语音设备的数量为0，则将所述拾音声强最高的语音设备作为应答语音设备。
在一个实施例中，所述第一计算单元具体用于：采用第二计算式计算该语音设备对应的唤醒评分，所述第二计算式包括：
当r在第一范围内时，P2＝a*s+b/|r-90|
当r在第二范围内时，P2＝b*s-a*|r-90|+c
式中，P2为所述唤醒评分；s为所述拾音声强；r为所述拾音角度；所述第一范围为：r大于等于60且小于80，或者，r大于100且小于等于120；所述第二范围为：r大于等于80且小于等于100；a和b为预设权重，且a大于b，c为10*a+b/10。
可理解的是，第二方面提供的装置和第一方面提供的方法是对应的，本方面中有关内容的解释、举例、有益效果等内容可以参见第一方面中的有关内容，此处不再赘述。
第三方面，本发明实施例提供一种基于协同纠错的唤醒系统，参见图8，包括在一个家庭内分布设置的语音设备以及与各个语音设备通信连接的云平台；所述语音设备中包括智能家电，所述智能家电中的语音模块具有语音拾取和语音反馈功能；所述云平台用于对全空间内的语音设备进行控制；所述决策设备为预先从所述语音设备中选出的一个智能家电，所述决策设备具有第二方面提供的基于协同纠错的唤醒装置。
可理解的是，该唤醒系统包括各个语音设备、云平台，当然还可以包括移动终端上的应用程序，该应用程序可以对各个语音设备进行控制。图8中的N为大于2的正整数。
可理解的是，第三方面提供的系统中有关内容的解释、举例、有益效果等内容可以参见第一方面、第二方面中的有关内容，此处不再赘述。
第四方面，本发明实施例提供一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现第一方面提供的方法的步骤。
具体地，可以提供配有存储介质的系统或者装置，在该存储介质上存储着实现上述实施例中任一实施例的功能的软件程序代码，且使该系统或者装置的计算机(或CPU或MPU)读出并执行存储在存储介质中的程序代码。
在这种情况下，从存储介质读取的程序代码本身可实现上述实施例中任何一项实施例的功能，因此程序代码和存储程序代码的存储介质构成了本发明的一部分。
用于提供程序代码的存储介质实施例包括软盘、硬盘、磁光盘、光盘(如CD-ROM、CD-R、CD-RW、DVD-ROM、DVD-RAM、DVD-RW、DVD+RW)、磁带、非易失性存 储卡和ROM。可选择地，可以由通信网络从服务器计算机上下载程序代码。
此外，应该清楚的是，不仅可以通过执行计算机所读出的程序代码，而且可以通过基于程序代码的指令使计算机上操作的操作系统等来完成部分或者全部的实际操作，从而实现上述实施例中任意一项实施例的功能。
此外，可以理解的是，将由存储介质读出的程序代码写到插入计算机内的扩展板中所设置的存储器中或者写到与计算机相连接的扩展模块中设置的存储器中，随后基于程序代码的指令使安装在扩展板或者扩展模块上的CPU等来执行部分和全部实际操作，从而实现上述实施例中任一实施例的功能。
可理解的是，第四方面提供的介质中有关内容的解释、举例、有益效果等内容可以参见第一方面、第二方面中的有关内容，此处不再赘述。
第五方面，本发明实施例提供一种语音设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如第一方面提供的所述方法的步骤。
可理解的是，该语音设备即决策设备。
可理解的是，第五方面提供的语音设备中有关内容的解释、举例、有益效果等内容可以参见第一方面、第二方面中的有关内容，此处不再赘述。
第六方面，为提高智能家电的智能联动运行，本发明实施例还提供了一种成套家电控制系统，该成套家电控制系统，包括物联网平台1、内容平台2、语义平台3、云平台4、语音获取模块7、智能终端5，物联网平台1可以为智汇家AIOT平台，该物联网平台具有配置模块11和智能控制场景模块12，其中，配置模块11和智能控制场景模块12通过通信网络与智能终端5进行信息传输并控制智能终端5的运行；云平台4分别与语义平台3、物联网平台1通过通信网络进行信息传输；移动端6通过无线的方式与云平台1实现信息传输；语义平台3包括语音解析模块33、功能反馈模块32和交互反馈模块31；语义平台3能够解析语音获取模块7获取的语音信号，并根据解析的语音信号输出控制指令，该控制指令通过无线通信的方式与云平台和物联网平台进行通信，进一步控制或驱动智能终端的正常运行。
进一步地，为实现远程控制和移动端控制，该种成套家电控制系统包括移动端6和语音播报模块8，移动端可以通过APP与物联网平台进行通信，或移动端6与智汇家AIOT平台进行通信，控制智能终端的正常运行；语音播报模块8完成语音播报。
进一步地，上述成套家电系统的语音指令系统包括语音拾取模块71、解析语音指令模块34、逻辑判断模块9、指定设备模块91、非指定设备模块92、AI分级管理模块311、AI智能分组模块312、语音播报模块81、执行智能设备指令模块10和反馈指令执行结果模块11；当语音拾取模块获取到用户的语音指令后，将语音指令传送到解析语音指令模块33，然后通过逻辑判断模块9的判断结果驱动控制智能终端运行，智能终端将语音指令的执行结果反馈并播报；其中AI分级管理模块311能够进行分级管理，AI智能分组模块312可以实现智能分组；其中指定设备模块91可以对特定的智能终端进行指定管理控制，非指定设备模块92可以对非特定的智能终端进行统一的管理控制。其中，指定的设备可以通过空间位置(如房间位置)和设备名称(空调或冰箱等)进行设备指定的定义。逻辑判断模块能够根据优先级对智能终端进行控制；AI分级管理模块能够根据智能终端的功能对空调、冰箱、洗衣机和电视的功能进行分类管理，该分类可以分为通用技能、专属技能和公共技能；AI智能分组模块能够根据空间区域(客厅、卧室、书房、厨房、阳台)对该空间区域内的智能终端(如空调、冰箱、洗衣机和电视进行归组)管理和控制。
可理解的是，语音拾取模块71和语音获取模块可以为一个模块，当然也可以不是同一个模块。解析语音指令模块34和语音解析模块33可以为同一个模块，当然也可以不是同一个模块。语音播放模块8和语音播放模块81可以为一个模块，当然也可以不是同 一个模块。
本说明书中的各个实施例均采用递进的方式描述，各个实施例之间相同相似的部分互相参见即可，每个实施例重点说明的都是与其他实施例的不同之处。尤其，对于装置实施例而言，由于其基本相似于方法实施例，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
本领域技术人员应该可以意识到，在上述一个或多个示例中，本发明所描述的功能可以用硬件、软件、挂件或它们的任意组合来实现。当使用软件实现时，可以将这些功能存储在计算机可读介质中或者作为计算机可读介质上的一个或多个指令或代码进行传输。
以上所述的具体实施方式，对本发明的目的、技术方案和有益效果进行了进一步详细说明，所应理解的是，以上所述仅为本发明的具体实施方式而已，并不用于限定本发明的保护范围，凡在本发明的技术方案的基础之上，所做的任何修改、等同替换、改进等，均应包括在本发明的保护范围之内。

Claims (18)

  1. 一种基于协同纠错的唤醒方法，其特征在于，在一个家庭的全空间内分布有语音设备，所述语音设备中包括智能家电，所述智能家电中的语音模块具有语音拾取和语音反馈功能；所述语音设备均与云平台通信连接；所述方法由决策设备执行，所述决策设备为预先从所述语音设备中选出的一个智能家电；所述方法包括：
    在接收到一个语音设备发送来的唤醒请求时，判断该语音设备所在的空间分组内是否存在其它语音设备；其中，一个处于唤醒等待状态的语音设备监听到唤醒词后向所述决策设备发送唤醒请求，一个家庭的全空间内的各个语音设备具有相同的唤醒词；
    若存在其它语音设备，则开始计时，在计时时长达到预设时长后判断在计时过程中是否接收到所述其它语音设备发送来的唤醒请求；
    若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送所述唤醒请求的语音设备中确定一个应答语音设备，并向所述应答语音设备发送唤醒应答指令，以使所述应答语音设备从唤醒等待状态进入唤醒状态。
  2. 根据权利要求1所述的方法，其特征在于，基于协同纠错的唤醒方法为协同纠错的就近唤醒方法，所述若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送所述唤醒请求的语音设备中确定一个应答语音设备，包括：基于就近原则从发送所述唤醒请求的语音设备中确定一个应答语音设备。
  3. 根据权利要求1所述的方法，其特征在于，若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送所述唤醒请求的语音设备中确定一个应答语音设备，包括：
    若在计时过程中接收到其它语音设备发送来的唤醒请求，则获取每一个发送唤醒请求的语音设备的拾音参数；
    根据发送唤醒请求的各个语音设备各自的拾音参数，从发送所述唤醒请求的语音设备中确定所述应答语音设备。
  4. 根据权利要求1所述的方法，其特征在于，各个语音设备均由所在家庭的用户移动终端上的同一个预设应用程序控制；所述决策设备由所述预设应用程序预先确定；其中，所述预设应用程序确定所述决策设备的过程包括：
    获取接入家庭网络的各个语音设备；
    获取加入所述家庭网络的每一个语音设备的带电状态数据、运算能力数据和用户使用习惯数据；
    根据每一个语音设备的带电状态数据、运算能力数据和用户使用习惯数据，确定每一个语音设备的决策能力评分；
    将所述决策能力评分最高的语音设备作为所述决策设备。
  5. 根据权利要求4所述的方法，其特征在于，所述根据每一个语音设备的带电状态数据、运算能力数据和用户使用习惯数据，确定每一个语音设备的决策能力评分，包括：采用第一计算式计算每一个语音设备的决策能力评分，所述第一计算式包括：
    P1＝u*(d2-1/y)
    式中，P1为所述决策能力评分；d为语音设备在过去一个月内的平均每天带电时长；y为语音设备的CPU运算能力；u为家庭用户在当前季节对语音设备的使用标记；若家庭用户在当前季节使用该语音设备，则该语音设备对应的使用标记为1；若家庭用户在当前季节不适用该语音设备，则该语音设备对应的使用标记为0。
  6. 根据权利要求4所述的方法，其特征在于，所述预设应用程序确定所述决策设备的过程还包括：所述预设应用程序通过心跳指令的方式按照预设时间间隔检测当前的决策设备是否处于带电状态；若未处于带电状态，则选择处于带电状态的各个语音设备中决策能力评分最高的一个语音设备作为当前的决策设备。
  7. 根据权利要求4所述的方法，其特征在于，所述空间分组为按照物理空间分组的方式对各个房间内的语音设备进行划分后得到的多个组，一个房间对应一个空间分组，一个空间分组内包括至少一个语音设备；
    对应的，所述预设应用程序用于提供配置界面，以使用户对接入家庭网络的每一个语音设备配置对应的房间标识；对应的，所述判断该语音设备所在的空间分组内是否存在其它语音设备，包括：从所述预设应用程序中获取该语音设备对应的房间标识，将该房间标识作为第一房间标识；判断其它语音设备的房间标识中是否存在与所述第一房间标识相同的房间标识；若存在，则该语音设备所在的房间内存在其它语音设备；否则，该语音设备所在的房间内不存在其它语音设备。
  8. 根据权利要求3所述的方法，其特征在于，所述拾音参数包括拾音角度和拾音声强；对应的，所述根据发送唤醒请求的各个语音设备各自的拾音参数，确定应答语音设备，包括：
    针对发送唤醒请求的各个语音设备，判断拾音角度在60～120度范围内的语音设备的数量是否大于1；
    若拾音角度在60～120度范围内的语音设备的数量大于1，则根据每一个拾音角度在60～120度范围内的语音设备的拾音角度和拾音声强，计算该语音设备对应的唤醒评分；一个语音设备的唤醒评分用于表征用户唤醒该语音设备的概率；将唤醒评分最高的语音设备作为应答语音设备；
    若所述拾音角度在60～120度范围的语音设备的数量为1，则将所述拾音角度在60～120度范围的语音设备作为应答语音设备；
    若所述拾音角度在60～120度范围的语音设备的数量为0，则将所述拾音声强最高的语音设备作为应答语音设备。
  9. 根据权利要求8所述的方法，其特征在于，所述计算该语音设备对应的唤醒评分，包括：采用第二计算式计算该语音设备对应的唤醒评分，所述第二计算式包括：
    当r在第一范围内时，P2＝a*s+b/|r-90|
    当r在第二范围内时，P2＝b*s-a*|r-90|+c
    式中，P2为所述唤醒评分；s为所述拾音声强；r为所述拾音角度；所述第一范围为：r大于等于60且小于80，或者，r大于100且小于等于120；所述第二范围为：r大于等于80且小于等于100；a和b为预设权重，且a大于b，c为10*a+b/10。
  10. 根据权利要求1～9任一项所述的方法，其特征在于，所述方法还包括如下至少一项：
    若发送唤醒请求的语音设备所在的空间分组内不存在其它语音设备，则向发送唤醒请求的语音设备发送唤醒应答指令，以使语音设备进入唤醒状态；
    若在计时过程中没有接收到其它语音设备发送来的唤醒请求，则确定发送唤醒请求的语音设备为误触发，则向发送唤醒请求的语音设备发送终止唤醒应答指令，以使发送唤醒请求的语音设备继续保持在唤醒等待状态。
  11. 根据权利要求1所述的方法，其特征在于，所述空间分组为按照发出唤醒词的用户在所述全空间内的位置和朝向进行划分后得到的虚拟空间分组，一个虚拟空间分组内包括至少两个语音设备；所述空间分组为所述云平台预先确定；
    其中，所述云平台确定空间分组的方法包括：所述云平台在历史唤醒任务中获取发出唤醒词的用户在所述全空间中的位置和朝向；获取各个语音设备拾取到所述预设唤醒词的声音强度，并根据所述声音强度从各个语音设备中选出第一语音设备；其中，各个所述第一语音设备所拾取到的声音强度高于其它的语音设备所拾取到的声音强度，且各个所述第一语音设备所拾取到的声音强度之间的最大差值在预设范围内；获取各个所述第一语音设备的拾音角度，并根据各个所述第一语音设备的拾音角度，从所述第一语音设备中选出拾音角度落在最佳拾音角度范围内的第二语音设备；将各个所述第二语音设备形成针对用户的所述位置和所述朝向的空间分组。
  12. 一种基于协同纠错的唤醒装置，其特征在于，在一个家庭的全空间内分布有语音设备，所述语音设备中包括智能家电，所述智能家电中的语音模块具有语音拾取和语 音反馈功能；各个所述语音设备均与云平台通信连接；
    所述装置安装在决策设备上，所述决策设备为预先从所述语音设备中选出的一个智能家电；所述装置包括：
    第一判断模块，用于在接收到一个语音设备发送来的唤醒请求时，判断该语音设备所在的空间分组内是否存在其它语音设备；其中，一个处于唤醒等待状态的语音设备监听到唤醒词后向所述决策设备发送唤醒请求，一个家庭的全空间内的各个语音设备具有相同的唤醒词；
    第二判断模块，用于若存在其它语音设备，则开始计时，在计时时长达到预设时长后判断在计时过程中是否接收到其它语音设备发送来的唤醒请求；
    目标确定模块，用于若在计时过程中接收到其它语音设备发送来的唤醒请求，从发送所述唤醒请求的语音设备中确定一个应答语音设备，并向所述应答语音设备发送唤醒应答指令，以使所述应答语音设备从唤醒等待状态进入唤醒状态。
  13. 一种基于协同纠错的唤醒系统，其特征在于，包括在一个家庭内分布设置的语音设备以及与各个语音设备通信连接的云平台；所述语音设备中包括智能家电，所述智能家电中的语音模块具有语音拾取和语音反馈功能；所述云平台用于对全空间内的语音设备进行控制；所述决策设备为预先从所述语音设备中选出的一个智能家电，所述决策设备上具有权利要求12所述的基于协同纠错的唤醒装置。
  14. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，该计算机程序被处理器执行时实现如权利要求1-11任一项所述方法的步骤。
  15. 一种语音设备，其特征在于，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如权利要求1-11任一项所述方法的步骤。
  16. 一种成套家电控制系统，其特征在于，包括物联网平台(1)、内容平台(2)、语义平台(3)、云平台(4)、语音获取模块(7)、智能终端(5)；其中，物联网平台(1)为智汇家AIOT平台，该物联网平台具有配置模块(11)和智能控制场景模块(12)，配置模块(11)和智能控制场景模块(12)通过通信网络与智能终端(5)进行信息传输并控制智能终端(5)的运行；云平台(4)分别与语义平台(3)和物联网平台(1)通过通信网络进行信息传输；语义平台(3)包括语音解析模块(33)、功能反馈模块(32)和交互反馈模块(31)，语义平台(3)能够解析语音获取模块(7)获取的语音信号，并根据解析的语音信号输出控制指令，该控制指令通过无线通信的方式分别与云平台(4)和物联网平台(1)进行通信，控制或驱动智能终端(5)的运行。
  17. 根据权利要求16所述的套家电控制系统，其特征在于，还包括移动端(6)和语音播报模块(8)，移动端(6)通过APP与物联网平台(1)进行通信，或移动端(6)与智汇家AIOT平台进行通信，控制智能终端(5)的运行，语音播报模块(8)用于语音播报。
  18. 根据权利要求16或17所述的套家电控制系统，其特征在于，还包括语音指令系统，该语音指令系统包括语音拾取模块(71)、解析语音指令模块(34)、逻辑判断模块(9)、指定设备模块(91)、非指定设备模块(92)、AI分级管理模块(311)、AI智能分组模块(312)、语音播报模块(81)、执行智能设备指令模块(10)和反馈指令执行结果模块(11)；当语音拾取模块获取到用户的语音指令后，将语音指令传送到解析语音指令模块(33)，然后通过逻辑判断模块(9)的判断结果驱动控制智能终端运行，智能终端将语音指令的执行结果反馈并播报；AI分级管理模块(311)用于进行分级管理，AI智能分组模块(312)用于实现智能分组；指定设备模块(91)用于对特定的智能终端进行指定管理控制，非指定设备模块(92)用于对非特定的智能终端进行统一的管理控制。
PCT/CN2023/096289 2022-05-31 2023-05-25 基于协同纠错的唤醒方法、装置及系统、介质、设备 WO2023231894A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210607070.8A CN115019793A (zh) 2022-05-31 2022-05-31 基于协同纠错的唤醒方法、装置及系统、介质、设备
CN202210607070.8 2022-05-31

Publications (1)

Publication Number Publication Date
WO2023231894A1 true WO2023231894A1 (zh) 2023-12-07

Family

ID=83070340

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/096289 WO2023231894A1 (zh) 2022-05-31 2023-05-25 基于协同纠错的唤醒方法、装置及系统、介质、设备

Country Status (2)

Country Link
CN (1) CN115019793A (zh)
WO (1) WO2023231894A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115019793A (zh) * 2022-05-31 2022-09-06 四川虹美智能科技有限公司 基于协同纠错的唤醒方法、装置及系统、介质、设备

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107622767A (zh) * 2016-07-15 2018-01-23 青岛海尔智能技术研发有限公司 家电系统的语音控制方法与家电控制系统
CN109391528A (zh) * 2018-08-31 2019-02-26 百度在线网络技术（北京）有限公司 语音智能设备的唤醒方法、装置、设备及存储介质
US20190392834A1 (en) * 2019-06-25 2019-12-26 Lg Electronics Inc. Method and apparatus for selecting voice-enabled device
CN111192591A (zh) * 2020-02-10 2020-05-22 百度在线网络技术（北京）有限公司 智能设备的唤醒方法、装置、智能音箱及存储介质
CN111640434A (zh) * 2020-06-05 2020-09-08 三星电子（中国）研发中心 用于控制语音设备的方法和装置
CN112037787A (zh) * 2020-10-20 2020-12-04 北京小米松果电子有限公司 唤醒控制方法、装置及计算机可读存储介质
CN113096658A (zh) * 2021-03-31 2021-07-09 歌尔股份有限公司 一种终端设备及其唤醒方法、装置和计算机可读存储介质
CN113689857A (zh) * 2021-08-20 2021-11-23 北京小米移动软件有限公司 语音协同唤醒方法、装置、电子设备及存储介质
CN114465837A (zh) * 2022-01-30 2022-05-10 云知声智能科技股份有限公司 一种智能语音设备协同唤醒处理方法及装置
CN115019793A (zh) * 2022-05-31 2022-09-06 四川虹美智能科技有限公司 基于协同纠错的唤醒方法、装置及系统、介质、设备

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107622767A (zh) * 2016-07-15 2018-01-23 青岛海尔智能技术研发有限公司 家电系统的语音控制方法与家电控制系统
CN109391528A (zh) * 2018-08-31 2019-02-26 百度在线网络技术（北京）有限公司 语音智能设备的唤醒方法、装置、设备及存储介质
US20190392834A1 (en) * 2019-06-25 2019-12-26 Lg Electronics Inc. Method and apparatus for selecting voice-enabled device
CN111192591A (zh) * 2020-02-10 2020-05-22 百度在线网络技术（北京）有限公司 智能设备的唤醒方法、装置、智能音箱及存储介质
CN111640434A (zh) * 2020-06-05 2020-09-08 三星电子（中国）研发中心 用于控制语音设备的方法和装置
CN112037787A (zh) * 2020-10-20 2020-12-04 北京小米松果电子有限公司 唤醒控制方法、装置及计算机可读存储介质
CN113096658A (zh) * 2021-03-31 2021-07-09 歌尔股份有限公司 一种终端设备及其唤醒方法、装置和计算机可读存储介质
CN113689857A (zh) * 2021-08-20 2021-11-23 北京小米移动软件有限公司 语音协同唤醒方法、装置、电子设备及存储介质
CN114465837A (zh) * 2022-01-30 2022-05-10 云知声智能科技股份有限公司 一种智能语音设备协同唤醒处理方法及装置
CN115019793A (zh) * 2022-05-31 2022-09-06 四川虹美智能科技有限公司 基于协同纠错的唤醒方法、装置及系统、介质、设备

Also Published As

Publication number Publication date
CN115019793A (zh) 2022-09-06

Similar Documents

Publication Publication Date Title
CN113516979B (zh) 在语音接口设备处的服务器提供的视觉输出
EP3535753B1 (en) Focus session at a voice interface device
CN108667697B (zh) 语音控制冲突解决方法、装置及语音控制系统
WO2019205134A1 (zh) 智能家居语音控制方法、装置、设备和系统
WO2020199673A1 (zh) 控制家电设备的方法、装置和家电设备
WO2023231894A1 (zh) 基于协同纠错的唤醒方法、装置及系统、介质、设备
WO2021047248A1 (zh) 基于多控制终端的iot设备控制方法、控制终端及存储介质
CN113096658A (zh) 一种终端设备及其唤醒方法、装置和计算机可读存储介质
CN110808886A (zh) 一种语音设备的控制方法、装置、存储介质及语音设备
CN111477230A (zh) 一种智能音箱系统、智能音箱系统的控制方法及存储介质
WO2021082131A1 (zh) 空气调节设备、温度控制方法和装置
CN108932942A (zh) 一种实现智能音箱人机对话的系统及其方法
CN110632854A (zh) 语音控制方法及装置、语音控制节点及系统及存储介质
WO2019227370A1 (zh) 一种多语音助手控制方法、装置、系统及计算机可读存储介质
CN111935517A (zh) 一种智能机顶盒的唤醒方法、装置、电子设备及存储介质
CN112585675A (zh) 选择地使用多个语音数据接收装置进行智能服务的方法、装置和系统
CN113138559A (zh) 设备交互方法、装置、电子设备及存储介质
CN116582382B (zh) 智能设备控制方法、装置、存储介质及电子设备
TWI831902B (zh) 聲控方法及終端裝置
CN114898750B (zh) 基于协同响应的智能家电控制方法、装置及系统、设备
CN112164398B (zh) 语音设备及其唤醒方法、装置以及存储介质
CN108663942A (zh) 一种语音识别设备控制方法、语音识别设备和中控服务器
CN112216279A (zh) 语音传输方法、智能终端及计算机可读存储介质
CN114815635A (zh) 计算机可读存储介质、智能面板及其语音交互方法
CN114999484A (zh) 交互语音设备的选举方法及系统

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: 23815077

Country of ref document: EP

Kind code of ref document: A1