WO2019007247A1 - Human-machine conversation processing method and apparatus, and electronic device - Google Patents

Abstract

Provided are a human-machine conversation processing method and apparatus, and an electronic device. The method comprises: after a device completing the last voice command, identifying the content of the last voice command to determine whether a user has the need to input the voice command again based on the last voice command; if it is determined that the user has the need to input a voice command again, initiating voice activity detection; otherwise, ending the conversation. The solution of the embodiment of the present invention can meet the requirement that a user wants to actively have continuous conversation with a device without repeatedly waking up the device, thereby improving the user experience, and improving the conversation efficiency.

Description

Manipine session processing method, device and electronic device

The present application claims priority to Chinese Patent Application No. JP-A No. No. No. No. No. No. No. No. No. No. No. No. No. .

Technical field

The present application relates to the field of computer technologies, and in particular, to a method, an apparatus, and an electronic device for processing a human-machine session.

Background technique

In the scenario of a human-machine session, after the user inputs a voice command to the device, the device executes the voice command input by the user, such as increasing the volume, playing the video, etc. After the instruction is executed, the text can also be passed through TTS (Text To Speech). Turn voice to feedback to the user, such as playing "Volume has increased", "Video is already on", etc. When the device completes a voice command and assumes that the entire session has terminated, it enters a sleep state.

However, in a continuous session scenario, if the user has further voice commands to input, the device needs to be re-awakened. Re-waking up the device will lead to greater inconvenience in both time and program. For example, the user has to re-enter the voice wake-up words, and the wake-up device will take a certain amount of time, thus seriously affecting the experience.

Summary of the invention

The invention provides a method, a device and an electronic device for processing a human-machine session. On the basis that the device does not need to be repeatedly awake, the user needs to actively perform continuous conversation with the device, improve the user experience and improve the session efficiency.

In order to achieve the above object, embodiments of the present invention adopt the following technical solutions:

In a first aspect, a method for processing a human-machine session is provided, including:

After the device completes the previous voice command, it identifies the content of the previous voice command to determine whether the user has the need to input the voice command again based on the previous voice command;

If it is determined that the user has a need to input a voice command again, voice activity detection is initiated; otherwise, the session is ended.

In a second aspect, another method for processing a human-machine session is provided, including:

Identifying the content of the received voice command;

Determining whether the user has a need to input a voice command again;

According to the judgment result, the human-machine session operation is performed.

The third aspect provides a processing device for a human-machine session, including:

The instruction identification module is configured to identify, after the device completes the previous voice instruction, the content of the previous voice instruction, to determine whether the user has the requirement to input the voice instruction again based on the previous voice instruction;

The voice detection module is configured to initiate voice activity detection if it is determined that the user has the requirement to input the voice command again; otherwise, the session is ended.

In a fourth aspect, a device for processing a human-machine session is provided, including:

a content identification module, configured to identify content of the received voice instruction;

a demand judging module, configured to determine whether the user has a need to input a voice command again;

The operation module is executed to perform a human-machine session operation according to the judgment result.

In a fifth aspect, an electronic device is provided, including:

Memory for storing programs;

a processor coupled to the memory for executing the program for:

After the device completes the previous voice command, it identifies the content of the previous voice command to determine whether the user has the need to input the voice command again based on the previous voice command;

If it is determined that the user has a need to input a voice command again, voice activity detection is initiated; otherwise, the session is ended.

In a sixth aspect, another electronic device is provided, including:

Memory for storing programs;

a processor coupled to the memory for executing the program for:

Identifying the content of the received voice command;

Determining whether the user has a need to input a voice command again;

According to the judgment result, the human-machine session operation is performed.

The method, device and electronic device for processing a human-machine session provided by the present invention can improve the device to execute continuous voice commands of the user by performing prediction and judgment on whether the user inputs the next voice command after the device completes the previous voice command. Efficiency and enhance the user experience.

The above description is only an overview of the technical solutions of the present application, and the technical means of the present application can be more clearly understood, and can be implemented in accordance with the contents of the specification, and the above and other objects, features and advantages of the present application can be more clearly understood. The following is a specific embodiment of the present application.

DRAWINGS

Various other advantages and benefits will become apparent to those skilled in the art from a The drawings are only for the purpose of illustrating the preferred embodiments and are not intended to be limiting. Throughout the drawings, the same reference numerals are used to refer to the same parts. In the drawing:

1 is a schematic diagram 1 of a process of a human-machine session according to an embodiment of the present invention;

2 is a second schematic diagram of processing of a human-machine session according to an embodiment of the present invention;

3 is a system structural diagram of processing of a human-machine session according to an embodiment of the present invention;

4a is a flowchart 1 of a method for processing a human-machine session according to an embodiment of the present invention;

4b is a flowchart 1 of a method for processing a human-machine session according to an embodiment of the present invention;

FIG. 5 is a structural diagram 1 of a processing apparatus for a human-machine session according to an embodiment of the present invention; FIG.

FIG. 5b is a second structural diagram of a device for processing a human-machine session according to an embodiment of the present invention;

6 is a third structural diagram of a device for processing a human-machine session according to an embodiment of the present invention;

FIG. 7 is a schematic structural diagram 1 of an electronic device according to an embodiment of the present invention; FIG.

FIG. 8 is a schematic structural diagram 2 of an electronic device according to an embodiment of the present invention.

Detailed ways

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While the embodiments of the present invention have been shown in the drawings, the embodiments Rather, these embodiments are provided so that this disclosure will be more fully understood and the scope of the disclosure will be fully disclosed.

In the scenario of an existing human-machine session, after the user inputs a voice command to the machine, the machine performs a voice command input by the user, such as increasing the volume, playing a video, etc., after the instruction is executed, the user can also feed back through the TTS. For example, the play "Volume has increased". Of course, feedback to the user through the TTS is not a necessary operation, and the device can perform voice feedback; or the device will respond to the voice command, for example, if the user inputs “speak a joke”, the device will select a joke to answer through the TTS. For example, if the user inputs "how is the weather today", the device will broadcast the weather forecast through TTS. For convenience of description, whether the device answers the question to the user through TTS or feedback or only performs the operation corresponding to the voice command without feedback, we refer to these situations as “device completion voice command”.

The invention changes the processing flow of ending the session immediately after the device completes the voice instruction in the prior art, and the core idea is that after the device completes the voice instruction, the content of the voice command input by the previous user is first determined and confirmed. Whether the user will also input the next voice command. If it is determined that the user also inputs the next voice command, after the device completes the voice command, it enters the voice activity detection (VAD) process, and if it is determined that the user does not input The next voice command terminates the session, thereby increasing the efficiency of the user's continuous session.

FIG. 1 is a schematic diagram of a process of a human-machine session according to an embodiment of the present invention. In the logic diagram, the basic flow of the human-machine session is user wake-up, VAD, voice input, ASR (Automatic Speech Recognition), semantic analysis, instruction execution, system feedback, and TTS. This process forms a closed loop. If the user needs to have a continuous voice command after judging the content of the previous voice command, the VAD can be restarted after the TTS to perform voice detection, and the original session flow is repeated.

In addition, the present application also solves the problem of judging the termination of the entire session in each link in the above process in consideration of the continuous session scenario.

The existing scheme for judging the termination of the session is the completion of the entire process from the user waking to the final TTS feedback as the session termination judgment condition, basically does not consider the continuous session scenario, and the error type judgment is performed once an abnormal situation occurs in each link in the session flow. And the abnormal feedback is performed through the TTS, and the TTS broadcasts that a session unit is terminated. In some special cases, such as when the voice system actively asks the user, the TTS will resume the session process from the VAD link after the broadcast.

2 is a processing logic diagram of terminating a session in a continuous session scenario. In this logic, there are generally five steps: domain judgment, VAD, ASR, semantic parsing, and execution of voice instructions.

Domain Judgment: Determine whether you need to enter the continuous session (continuous session) state after the previous voice command is completed. If it is determined that the user has a need to further issue a voice command according to the content of the previous voice command, the VAD is started; otherwise, the session is ended.

VAD: After the VAD is started, if a voice signal is detected within the set time, the voice signal is sent to the ASR for voice analysis to form a text; if no voice signal is detected, the session is ended.

ASR: Perform text parsing on the speech signal. If the text content is parsed, the text content is semantically parsed; if the text content is not parsed after the text parsing, the session is terminated.

Semantic parsing: Semantic parsing of text content, judging whether the statement in the text enters the preset field, if the statement does not enter any field, or the statement enters the field of stopping the radio (terminating session), the session is terminated; if the statement When entering an existing field, voice instructions are formed according to the existing field.

Execution command: according to the determined voice instruction, control the corresponding device to execute the voice instruction and feed back through TTS.

After the execution of the voice instruction is completed, the flow redirects to the first step, that is, to continue to judge the field of the previous instruction to determine whether the user needs a continuous session.

Based on the schematic diagram of the processing method of the human-machine session shown in FIG. 1 and FIG. 2, the embodiment of the present invention provides a processing system for the human-machine session, which is used to improve the efficiency of the user's active continuous session in the human-machine session scenario. As shown in FIG. 3, the system includes: a device 310 and a server 320.

Apparatus 310 includes:

Human-computer interaction devices in the human-machine session, such as microphones, stereos, etc., and operating devices that execute voice commands, such as media playback devices, air conditioners, televisions, refrigerators, and the like.

The device 310 is configured to interact with a person during a human-machine session, including voice signal collection, TTS feedback, and the like, and perform specific operations of the voice instruction.

The server 2 has logic processing functions for controlling VAD startup, ASR, semantic analysis, forming voice control commands, and feeding back to the device.

As shown in Figure 3, the server 2 specifically includes: a human machine session processing device 321 and a domain library 322;

The processing device 321 of the human-machine session includes:

The instruction identification module is configured to: after the device 310 completes the previous voice instruction, identify the content of the previous voice instruction, determine whether the user has the requirement to input the voice instruction again based on the previous voice instruction; and perform the content of the voice instruction At the time of recognition, the voice instruction needs to be judged in the domain library 322 in the demand type. A plurality of intents in a plurality of different fields are preset stored in the domain library 322.

The so-called domain is a function that implements a certain type of user's needs in a human-computer interaction system. Domain identification is the process of determining which type of demand a user's voice belongs to.

Intention, the function of realizing a single explicit requirement of a user in a certain field in a human-computer interaction system. Intent recognition is the process of determining which of a user's voice belongs to a specific requirement in a certain field.

By identifying the content of the user's previous voice command in the domain library 322, it can be determined whether the user still has a need to re-enter the voice command.

The voice detection module is configured to start the voice activity detection VAD if it is determined that the user has the requirement to input the voice instruction again; otherwise, end the session.

On the side of the device 310, the device for receiving the voice signal, such as the microphone, is always on, but only after the server 320 determines the detection flow for starting the VAD, the voice signal received by the microphone is transmitted to the processing device 321 of the human-machine session. As a voice command signal in the human-machine session flow. The VAD detection process is only started at the beginning of each session to detect the voice signal. When the system considers that there is no voice input, the VAD is automatically turned off until the user wakes up the device again to start the VAD again. Therefore, when the command recognition module determines that the user has a need to input the voice command again, the voice detection module is triggered to activate the VAD.

The voice detection module is further configured to end the session if the voice signal is not detected within the specified detection time after starting the VAD; otherwise, the voice recognition module is triggered to perform automatic voice recognition ASR on the detected voice signal.

Time calculation module for:

Counting the first average time that the user wakes up from the device to the voice command during each session; and counts the second average time that the user sends the voice command from the VAD to the user during each session; The average time and the second average time are calculated to obtain a specified detection time.

The time of VAD detection has a greater impact on the effect of VAD. The detection time of the VAD should be determined according to the pronunciation time of the user in the continuous conversation scenario, and the pronunciation habits of different users are different. This requires dynamic adjustment of the VAD execution time, specifically based on the user's habit of using the speaker to dynamically optimize the VAD detection time. The time counted by these habits includes the average time the user wakes up from the device to the time the voice command is issued, and the average time from the start of the VAD to the user's voice command.

Further, the time calculation module calculates the specified detection time according to the first average time and the second average time, and may include:

According to T4=T3+(T2-T3)/2

The specified detection time T4 is calculated; wherein T2 is the sum of the first average time T1 and the preset redundancy time, and T3 is the second average time.

The speech recognition module is further configured to: after the ASR is performed on the detected speech signal, if the text content is not recognized, the session is ended; otherwise, the semantic parsing module is triggered to perform semantic analysis on the recognized text content.

The semantic parsing module is configured to end the session after the semantic analysis of the recognized text content, if the parsed semantics does not enter any of the preset fields, or the parsed semantics are explicitly ended in the session; Otherwise, the voice instruction is generated according to the domain entered by the parsed semantics, and the instruction execution module is triggered to control the corresponding device to perform the operation according to the voice instruction.

The processing system for the human-machine session provided by the embodiment of the present invention identifies the content of the previous voice command after the device completes the previous voice command, and determines whether the user has the requirement to input the voice command again based on the previous voice command; If it is determined that the user has the need to input the voice command again, the voice activity detection VAD is started; otherwise, the session is ended. Further, in the continuous session process, according to the execution status of each link, the conditions for ending the session are set in advance, and after the determination condition is formed, the session is ended, and the complete process of the continuous session is realized.

The technical solutions of the present application are further described below through various embodiments.

Embodiment 1

The flowchart of the method for determining whether to perform a continuous session according to the content of the voice instruction is as shown in FIG. 4a, which is a flow chart 1 of the processing method of the human-machine session shown in the embodiment of the present invention, and the execution body of the method is shown in FIG. The processing device of the human-machine session. As shown in FIG. 4a, the processing method of the human-machine session includes the following steps:

S410: After the device completes the previous voice command, identify the content of the previous voice command, and determine whether the user has the requirement to input the voice command again based on the previous voice command.

In the existing human-machine session scenario, after inputting a voice command, the user often wants to input a new voice command again based on the content of the voice command and the execution result. For example, the previous voice input is “helping me to search for a deformation. King Kong", after the system search is completed, the search result list is displayed through the screen. At this time, the user is likely to input the voice command of "playing the first" search result.

In order to improve the efficiency of the human-machine session on the premise of satisfying the continuous session requirement of the user, in this embodiment, after completing a voice command, the content of the previous completed voice command is identified to determine whether the user has a basis. The requirement to input a voice command again with a voice command. Thereby determining whether to initiate the process of a continuous session.

Domain judgment, that is, whether to determine the need to enter a continuous session state after a domain command is completed. The triggering of a continuous session needs to be determined according to the type of the user's previous voice command and the execution status.

The basis for judging a continuous conversation comes from but is not limited to the following principles:

1. Is it clear that continuous sessions are not needed; for example, the control class instruction ("I want to turn the light on") is basically a user's operation, and no continuous session is required in such a scenario;

2. Is there often a contextual appeal: such as the weather area ("How is the weather today?" - "Tomorrow");

3. Is it a continuous multi-instruction requirement: such as a movie query ("Help me search for Transformers" - "play the first one" - "full screen playback").

Based on the above principles, it is determined whether a continuous session is required. If a continuous session is required, the user is considered to have a need to input the voice command again based on the previous voice command.

If it is determined that the user has a need to input the voice instruction again, step S420 is performed, otherwise, S430 is performed.

S420, starting a voice activity detection VAD;

S430, end this session.

When it is determined that the user has the requirement to input the voice command again, the device side can be controlled to open the VAD detection process to collect the voice command that the user may input, and upload it to the processing device of the human-machine session on the server side for identification processing.

When it is determined that the user does not need to input the voice command again, the session can be ended and the control device enters the standby state.

In the embodiment, the content of the previous voice command is identified, and the determination process of determining whether the user has the requirement to input the voice command again based on the previous voice command may be processed on the server side, that is, whether the front end device is turned on by the server. Read the session process.

Further, the above method further includes:

After the VAD is started, if no voice signal is detected within the specified detection time, the session is ended; otherwise, the detected voice signal is automatically voice-recognized ASR.

When it is determined that the user has the requirement to input the voice instruction again, the server side controls the device side to start the VAD process, and transmits the voice signal detected by the microphone to the server side for analysis. If the vocal input is not detected within the specified detection time, the session is considered to be ended and the VAD process is closed. If the vocal input is detected within the specified detection time, the detected sound is uploaded to the server for ASR processing.

Vocal detection needs to shield out noise interference. Steady-state noise is better to identify and shield, such as air-conditioning noise with stable frequency and motor noise; but dynamic noise is more difficult to shield, such as singing, TV noise and other frequency changes. It also contains the noise of vocal recordings. Therefore, the detection time of VAD has a greater influence on the effect of VAD.

The detection time of the VAD should be determined according to the pronunciation time of the user in a continuous session, and the pronunciation habits of different users are different. The dynamic adjustment of the VAD is based on the user's habit of using the speaker dynamics to optimize the VAD detection time. The optimization strategy is as follows:

Counting the first average time that the user wakes up from the device to the voice command during each session; and counts the second average time that the user sends the voice command from the VAD to the user during each session; The average time and the second average time are calculated to obtain the specified detection time.

Specifically, during the session of the user, the average time T1 (first average time) from the wake-up of the device to the issuance of the command is calculated in real time in the scenario of the normal wake-up device.

Set the detection time of the VAD in the initial continuous session to T2=T1+3(s). Because the user's pronunciation is generally slower than the normal wake-up under the continuous session state, the detection time of the initial continuous session defaults to Add 3 seconds (redundancy time) based on the evaluation time in the state of the user wake-up device.

The detection time of the VAD in the initial continuous session is a long fault-tolerant time. It needs to be converged according to the actual average pronunciation time of the user in the continuous session state. In the course of each session of the user, the continuous session needs to be calculated in real time. In the state, the average time T3 (second average time) from the start of VAD to the user's voice command.

Based on the detection time T2 of the VAD in the initial continuous session, T2 is corrected by the actual average utterance time T3 of the user in the continuous session state, so that a relatively reasonable designated detection time T4 can be obtained.

Further, the processing step of calculating the specified detection time according to the first average time and the second average time may include:

By the following formula:

T4=T3+(T2-T3)/2 (1)

The specified detection time T4 is calculated; wherein T2 is the sum of the first average time T1 and the preset redundancy time, and T3 is the second average time.

Further, the above method further includes:

After the ASR is performed on the detected speech signal, if the text content is not recognized, the session is ended; otherwise, the recognized text content is semantically parsed.

There are still conditions for judging the termination of a continuous session during the ASR phase. When there is noise in the VAD link, the ASR may not recognize the voice input. At this time, the recognition result may return to null. Therefore, it can be considered that the session is returned to null when the ASR operation result is empty in the continuous session scenario. End. However, there is also a case where the user's voice command is not recognized by the objective factor, and the ASR cannot recognize that the return is empty. Therefore, when the session is terminated in the ASR session, the user equipment can be notified by the TTS that it is not clear, and wake up again if necessary.

When the execution of the ASR is completed, the text content is recognized, and the text content can be further transmitted to the next processing step, that is, semantic analysis.

Further, the above method further includes:

After the semantic analysis of the recognized text content, if the parsed semantics does not enter any of the preset fields, or the parsed semantics are explicitly ended to end the session, the session ends; otherwise, according to the parsing The field into which the semantics enter generates voice instructions and controls the corresponding device to perform operations according to the voice instructions.

The judgment of continuous session termination can be made by understanding the actual semantics of the user's voice instructions.

Semantic analysis can be divided into three parts, domain identification, intention identification, and execution logic judgment. Among them, the judgment to judge whether the continuous dialogue is terminated can be basically completed in the field identification link, which may include but not limited to the following two judgments. condition:

The session terminates when the semantics in the text parsed by ASR does not fall into any of the fields (multiple different fields are preset in the domain library). If the interference noise is misjudged into the semantic analysis stage in a continuous session scenario, the messy and non-logical statements are not easily understood by semantics into a certain domain, and the session is terminated.

When the parsed semantics falls into the "stop word" field, it indicates that the user explicitly uses the instruction to stop the continuous session, and the session is terminated. The corpus in the field of “stop words” is “good”, “thank you”, “nothing is wrong”...

When the semantics falls into other fields than the "stop word", the field instruction is normally executed, that is, the voice instruction is generated according to the specific intention in the field that falls into, and then the corresponding operation is performed according to the voice instruction control device.

After the instruction is successfully executed, the process returns to the initial step S410 to determine the domain of the previous voice command, determine whether the user has the need to input the voice command again based on the previous voice command, and start the VAD process.

If the instruction execution fails, step S410 may be skipped, the VAD is directly turned on, and the continuous dialogue scenario is entered from step S420.

The processing method of the human-machine session provided by the embodiment of the present invention, after the device completes the previous voice instruction, identifies the content of the previous voice command, and determines whether the user has the requirement to input the voice command again based on the previous voice command; If it is determined that the user has the requirement to input the voice instruction again, the voice activity detection is started; otherwise, the session is ended, thereby effectively improving the session processing efficiency in the continuous session scenario.

Further, in the continuous session execution process, by pre-making the judgment conditions of the current session for each link, the session process can be ended when the conditions are satisfied, and the integrity of the continuous session is ensured.

Embodiment 2

FIG. 4b illustrates another method for processing a human-machine session according to an embodiment of the present invention. The method is slightly changed based on the method shown in FIG. 4a. As shown in FIG. 4b, the processing method of the human-machine session includes the following steps:

S440, identifying the content of the received voice command.

In the existing human-machine session scenario, after inputting a voice command, the user often wants to input a new voice command again based on the content of the voice command and the execution result. For example, the previous voice input is “helping me to search for a deformation. King Kong", after the system search is completed, the search result list is displayed through the screen. At this time, the user is likely to input the voice command of "playing the first" search result.

In order to improve the efficiency of the human-machine session on the premise of satisfying the continuous session requirements of the user, in the embodiment, after the voice command is received by the device, the content of the voice command is first identified. For the specific identification process, refer to step S410. related information.

S450, determining whether the user has a need to input a voice instruction again;

After identifying the content of the voice command, it is determined whether there is a need to input the voice command again based on the voice command. For the specific judgment process, refer to the related content of step S410.

For example, the historical data may be collected for the result of whether the user performs voice input again for the personalized habit of different users (voiceprint recognition), and the probability that the user inputs the voice instruction again is calculated according to the statistical result. If the obtained probability is greater than the preset probability threshold, it is determined that the user has a need to input the voice instruction again; otherwise, the user is determined not to input the voice command again.

S460. Perform a human-machine session operation according to the judgment result.

Through the content of the voice command input by the user, it is determined whether the user has a need to input the voice again, and the subsequent processing operation is guided according to the judgment result. For example, if it is determined that the user has the requirement to input the voice instruction again, the start voice activity detection is performed after the device completes the voice command to implement the continuous session; otherwise, the session ends after the device completes the voice command.

In addition, it should be noted that the steps in the method shown in the first embodiment can also be performed in the method steps in this embodiment, and the step principle is not described herein.

The processing method of the human-machine session provided by the embodiment of the present invention identifies the content of the received voice command, determines whether the user has the requirement to input the voice command again; and then performs the human-machine session operation according to the judgment result, thereby effectively improving Session processing efficiency in a continuous session scenario.

Embodiment 3

As shown in FIG. 5, which is a structural diagram of a processing device for a human-machine session according to an embodiment of the present invention, the processing device of the human-machine session can be used to perform the method steps shown in FIG. 4a, including:

The instruction identification module 510 is configured to: after the device completes the previous voice instruction, identify the content of the previous voice instruction, and determine whether the user has the requirement to input the voice instruction again based on the previous voice instruction;

The voice detection module 520 is configured to start the voice activity detection VAD if it is determined that the user has the requirement to input the voice instruction again; otherwise, end the session.

Further, as shown in Figure 5b, the above apparatus further includes a voice recognition module 530;

The voice detection module 520 is further configured to: after the VAD is started, if the voice signal is not detected within the specified detection time, the session is ended; otherwise, the triggered voice recognition module 530 performs automatic voice recognition on the detected voice signal. .

Further, as shown in FIG. 5b, the foregoing apparatus further includes a time calculation module 540, configured to: collect a first average time that the user wakes up from the device to the voice command during each session, and counts the user in each time. During the session, the second average time from the initiation of the VAD to the user's voice command; the specified detection time is calculated according to the first average time and the second average time.

Further, the time calculation module 540 calculates the specified detection time according to the first average time and the second average time, including:

According to T4=T3+(T2-T3)/2

The specified detection time T4 is calculated; wherein T2 is the sum of the first average time T1 and the preset redundancy time, and T3 is the second average time.

Further, as shown in Figure 5b, the above apparatus further includes a semantic parsing module 550;

The speech recognition module 540 is further configured to: after the ASR is performed on the detected speech signal, if the text content is not recognized, the current session is ended; otherwise, the trigger semantic parsing module 550 performs semantic parsing on the recognized text content.

Further, as shown in Figure 5b, the above apparatus further includes an instruction execution module 560;

The semantic parsing module 550 is further configured to: after semantically parsing the recognized text content, if the parsed semantics does not enter any preset domain, or the parsed semantics are explicitly ended, the end is Session; otherwise, the voice instruction is generated according to the domain entered by the parsed semantics, and the trigger instruction execution module 560 controls the corresponding device to perform the operation according to the voice instruction.

The processing device for the human-machine session provided by the embodiment of the present invention identifies the content of the previous voice command after the device completes the previous voice command, and determines whether the user has the requirement to input the voice command again based on the previous voice command; If it is determined that the user has the requirement to input the voice command again, the voice activity detection VAD is started; otherwise, the session is ended, thereby effectively improving the session processing efficiency in the continuous session scenario.

Further, in the continuous session execution process, by pre-making the judgment conditions of the current session for each link, the session process can be ended when the conditions are satisfied, and the integrity of the continuous session is ensured.

Embodiment 4

As shown in FIG. 6, which is a structural diagram of a processing device for a human-machine session according to an embodiment of the present invention, the processing device of the human-machine session can be used to perform the method steps shown in FIG. 4b, including:

a content identification module 610, configured to identify content of the received voice instruction;

The requirement judging module 620 is configured to determine whether the user has a requirement for inputting a voice instruction again;

The operation module 630 is configured to perform a human-machine session operation according to the determination result.

The processing device for the human-machine session provided by the embodiment of the present invention identifies the content of the received voice command, determines whether the user has the requirement to input the voice command again, and then performs the human-machine session operation according to the determination result, thereby effectively improving Session processing efficiency in a continuous session scenario.

Embodiment 5

The foregoing embodiment 3 describes the overall architecture of the processing device of the human-machine session, and the function of the device can be implemented by using an electronic device, as shown in FIG. 7 , which is a schematic structural diagram of the electronic device according to the embodiment of the present invention, specifically including Memory 710 and processor 720.

The memory 710 is configured to store a program.

In addition to the above described procedures, memory 710 can also be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on an electronic device, contact data, phone book data, messages, pictures, videos, and the like.

Memory 710 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Disk or Optical Disk.

The processor 720 is coupled to the memory 710 for executing a program in the memory 710 for:

After the device completes the previous voice command, it identifies the content of the previous voice command to determine whether the user has the need to input the voice command again based on the previous voice command;

If it is determined that the user has a need to input a voice command again, the voice activity detection VAD is initiated; otherwise, the session is ended.

The specific processing operations described above have been described in detail in the foregoing embodiments, and are not described herein again.

Further, as shown in FIG. 7, the electronic device may further include: a communication component 730, a power component 740, an audio component 750, a display 760, and the like. Only some of the components are schematically illustrated in FIG. 7, and it is not meant that the electronic device includes only the components shown in FIG.

Communication component 730 is configured to facilitate wired or wireless communication between the electronic device and other devices. The electronic device can access a wireless network based on a communication standard such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, communication component 730 receives broadcast signals or broadcast associated information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, communication component 730 also includes a near field communication (NFC) module to facilitate short range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

A power component 740 provides power to various components of the electronic device. Power component 740 can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the electronic device.

The audio component 750 is configured to output and/or input an audio signal. For example, the audio component 750 includes a microphone (MIC) that is configured to receive an external audio signal when the electronic device is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal may be further stored in memory 710 or transmitted via communication component 730. In some embodiments, audio component 750 also includes a speaker for outputting an audio signal.

The display 760 includes a screen whose screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can sense not only the boundaries of the touch or sliding action, but also the duration and pressure associated with the touch or slide operation.

Embodiment 6

The foregoing embodiment 4 describes the overall architecture of the processing device of the human-machine session. The function of the device can be implemented by using an electronic device. As shown in FIG. 8 , it is a schematic structural diagram of the electronic device according to the embodiment of the present invention, which specifically includes Memory 810 and processor 820.

The memory 810 is configured to store a program.

In addition to the above described procedures, memory 810 can also be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on an electronic device, contact data, phone book data, messages, pictures, videos, and the like.

The memory 810 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable. Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Disk or Optical Disk.

The processor 820 is coupled to the memory 810 for executing a program in the memory 810 for:

Identifying the content of the received voice command;

Determining whether the user has a need to input a voice command again;

According to the judgment result, the human-machine session operation is performed.

The specific processing operations described above have been described in detail in the foregoing embodiments, and are not described herein again.

Further, as shown in FIG. 8, the electronic device may further include: a communication component 830, a power component 840, an audio component 850, a display 860, and the like. Only some of the components are schematically illustrated in FIG. 8, and it is not meant that the electronic device includes only the components shown in FIG.

Communication component 830 is configured to facilitate wired or wireless communication between the electronic device and other devices. The electronic device can access a wireless network based on a communication standard such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, communication component 830 receives broadcast signals or broadcast associated information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, communication component 830 also includes a near field communication (NFC) module to facilitate short range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

A power supply assembly 840 provides power to various components of the electronic device. Power component 840 can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the electronic device.

The audio component 850 is configured to output and/or input audio signals. For example, the audio component 850 includes a microphone (MIC) that is configured to receive an external audio signal when the electronic device is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal may be further stored in memory 810 or transmitted via communication component 830. In some embodiments, audio component 850 also includes a speaker for outputting an audio signal.

The display 860 includes a screen whose screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can sense not only the boundaries of the touch or sliding action, but also the duration and pressure associated with the touch or slide operation.

One of ordinary skill in the art will appreciate that all or part of the steps to implement the various method embodiments described above may be accomplished by hardware associated with the program instructions. The aforementioned program can be stored in a computer readable storage medium. The program, when executed, performs the steps including the foregoing method embodiments; and the foregoing storage medium includes various media that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.

Finally, it should be noted that the above embodiments are only for explaining the technical solutions of the present application, and are not limited thereto; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that The technical solutions described in the foregoing embodiments may be modified, or some or all of the technical features may be equivalently replaced; and the modifications or substitutions do not deviate from the technical solutions of the embodiments of the present application. range.

Claims (13)

1. A method for processing a human-machine session, comprising:

   After the device completes the previous voice command, it identifies the content of the previous voice command to determine whether the user has the need to input the voice command again based on the previous voice command;

   If it is determined that the user has a need to input a voice command again, voice activity detection is initiated; otherwise, the session is ended.
2. The method of claim 1 further comprising:

   After the voice activity detection is initiated, if no voice signal is detected within the specified detection time, the session is ended; otherwise, automatic voice recognition is performed on the detected voice signal.
3. The method of claim 2, wherein the method further comprises:

   Counting, during each session, the first average time that the user wakes up from the device to the voice command;

   Counting, by the user, a second average time from the initiation of the voice activity detection to the user's voice command during each session;

   The specified detection time is calculated according to the first average time and the second average time.
4. The method according to claim 3, wherein the calculating the specified detection time according to the first average time and the second average time comprises:

   According to T4=T3+(T2-T3)/2

   Calculating the specified detection time T4; wherein the T2 is a sum of the first average time T1 and a preset redundancy time, and the T3 is the second average time.
5. The method of claim 2, wherein the method further comprises:

   After the automatic speech recognition is performed on the detected speech signal, if the text content is not recognized, the session is ended; otherwise, the recognized text content is semantically parsed.
6. The method of claim 5, wherein the method further comprises:

   After semantically parsing the identified text content, if the parsed semantics does not enter any of the preset fields, or the parsed semantics are explicitly ending the session, the session ends; otherwise, according to the parsing The obtained semantic field generates a voice instruction, and controls the corresponding device to perform an operation according to the voice instruction.
7. A method for processing a human-machine session, comprising:

   Identifying the content of the received voice command;

   Determining whether the user has a need to input a voice command again;

   According to the judgment result, the human-machine session operation is performed.
8. The method according to claim 7, wherein the performing the human-machine session operation according to the determination result comprises:

   If it is determined that the user has the requirement to input the voice instruction again, the start voice activity detection is performed after the device completes the voice command; otherwise, the session ends after the device completes the voice command.
9. The method according to claim 7, wherein the determining whether the user has the voice command input again comprises:

   Calculate the probability that the user inputs the voice command again;

   If the probability is greater than the preset probability threshold, it is determined that the user has a need to input the voice instruction again; otherwise, the user is determined not to input the voice command again.
10. A processing device for a human-machine session, comprising:

    The instruction identification module is configured to identify, after the device completes the previous voice instruction, the content of the previous voice instruction, to determine whether the user has the requirement to input the voice instruction again based on the previous voice instruction;

    The voice detection module is configured to initiate voice activity detection if it is determined that the user has the requirement to input the voice command again; otherwise, the session is ended.
11. A processing device for a human-machine session, comprising:

    a content identification module, configured to identify content of the received voice instruction;

    a demand judging module, configured to determine whether the user has a need to input a voice command again;

    The operation module is executed to perform a human-machine session operation according to the judgment result.
12. An electronic device, comprising:

    Memory for storing programs;

    a processor coupled to the memory for executing the program for:

    After the device completes the previous voice command, it identifies the content of the previous voice command to determine whether the user has the need to input the voice command again based on the previous voice command;

    If it is determined that the user has a need to input a voice command again, voice activity detection is initiated; otherwise, the session is ended.
13. An electronic device, comprising:

    Memory for storing programs;

    a processor coupled to the memory for executing the program for:

    Identifying the content of the received voice command;

    Determining whether the user has a need to input a voice command again;

    According to the judgment result, the human-machine session operation is performed.

Images