A voice information processing method, device, system and storage medium

Abstract

The application discloses a voice information processing method, device and system and a storage medium, and aims to improve voice interaction effect. The method comprises the following steps: when voice information is received in a wake-up state, it is judged whether the voice information and a wake-up word received when the wake-up state is changed are issued by the same person; when the voice information and the wake-up word received when the wake-up state is changed are issued by the same person, the voice information is recognized; and an instruction corresponding to the voice information is executed according to a recognition result of the voice information. By adopting the scheme provided in the application, the probability of misrecognition is reduced, and the voice interaction effect is improved.

Description

Technical Field

[0001] This application relates to the field of voice information processing technology, and in particular to a voice information processing method, apparatus, system and storage medium. Background Technology

[0002] In many situations, multiple users may be present during voice interaction. If a user wants to interact via voice and activates the device with a wake-up word, the device may misinterpret the conversations of other users as commands. For example, in a car's voice interaction scenario, if multiple people are in the vehicle conversing, misrecognition is very likely to occur, resulting in poor voice interaction performance. Summary of the Invention

[0003] This application provides a voice information processing method, apparatus, system, and storage medium to improve the voice interaction effect.

[0004] This application provides a voice information processing method, including:

[0005] When a voice message is received in the wake-up state, it is determined whether the voice message and the wake-up word received when the wake-up state is changed are issued by the same person.

[0006] When the voice information is issued by the same person as the wake word received when the state is changed to wake-up, the voice information is identified.

[0007] Execute the instruction corresponding to the voice information based on the recognition result of the voice information.

[0008] The beneficial effects of this application are as follows: when receiving voice information, it determines whether the voice information and the wake word received when changing to the wake state are issued by the same person before recognizing the voice information and executing the instruction corresponding to the voice message based on the recognition result. In other words, in the wake state, only the voice instructions issued by the user who issued the wake word are executed, and the voice instructions issued by other users are not executed, which reduces the probability of misrecognition and improves the voice interaction effect.

[0009] In one embodiment, the method further includes:

[0010] When a voice message is received in a non-wake-up state, it is determined whether the voice message is a wake-up word;

[0011] When the voice information is a wake word, the current state is changed to a wake state, and the wake word is stored;

[0012] The step of determining whether the voice information and the wake-up word received when changing to the wake-up state are issued by the same person includes:

[0013] Determine whether the voice information and the wake word stored when changing the current state to the wake state were spoken by the same person.

[0014] In one embodiment, determining whether the voice information and the wake word stored when changing the current state to a wake-up state were spoken by the same person includes:

[0015] The voice information is purified;

[0016] The first voiceprint information of the purified speech information is compared with the second voiceprint information of the wake word stored when the current state is changed to the wake state to determine the similarity between the first voiceprint information and the second voiceprint information.

[0017] When the similarity between the first voiceprint information and the second voiceprint information is greater than a preset similarity, it is determined that the voice command and the wake-up word stored when changing the current state to the wake-up state were issued by the same person.

[0018] In one embodiment, the purification of the voice information includes:

[0019] Extract the audio frames corresponding to the voice information;

[0020] Determine the ratio of signal power to interference power for each audio frame;

[0021] Audio frames whose signal power to interference power ratio is less than a preset ratio are removed. The audio information composed of the remaining audio frames after removing audio frames whose power to interference power ratio is less than the preset ratio is the purified audio information.

[0022] In one embodiment, determining the ratio of signal power to interference power for each audio frame includes:

[0023] When multiple audio signals are detected in each audio frame, the audio signal with the highest power among the multiple audio signals is determined to be the valid signal, and the remaining audio signals are interference signals.

[0024] The power of the valid signal is determined as the signal power of each audio frame, and the power of the interference signal is determined as the interference power of each audio frame.

[0025] Determine the ratio of signal power to interference power for each audio frame.

[0026] In one embodiment, the step of discarding audio frames whose signal power to interference power ratio is less than a preset ratio includes:

[0027] Determine whether the frame length of each audio frame is the same as the speech frame length specified by the speech activity detection model;

[0028] When the frame length of each audio frame is different from the speech frame length specified by the speech activity detection model, a first ratio between the frame length of the audio frame and the speech frame length specified by the speech activity detection model is determined.

[0029] Determine a second ratio between the original index value of the speech activity detection model and the first ratio;

[0030] The multiple integers obtained by rounding down the second ratio are determined as new index values;

[0031] The original index value of the speech activity detection model is replaced with a new index value, wherein after the index value is replaced, the speech activity detection model reads audio frames according to the new index value, and the new index value corresponds to the position of the audio frame;

[0032] Each audio frame, and the ratio of signal power to interference power of each audio frame, are input into the speech activity detection model after the replacement index value. The speech activity detection model is used to purify audio frames by eliminating audio frames whose ratio of signal power to interference power is less than a preset ratio.

[0033] The purified audio frame output by the speech activity detection model after obtaining the replaced index value.

[0034] In one embodiment, the method further includes:

[0035] The voice command is refused to be executed if it is not issued by the same person as the wake word stored when the user changes to the wake-up state.

[0036] This application also provides a voice information processing device, including:

[0037] The first judgment module is used to determine whether the voice information received in the wake-up state is issued by the same person as the wake-up word received when the wake-up state is changed.

[0038] The recognition module is used to recognize the voice information when the voice information is issued by the same person as the wake word received when the state is changed to wake-up.

[0039] The execution module is used to execute the instructions corresponding to the voice information based on the recognition result of the voice information.

[0040] In one embodiment, the apparatus further includes:

[0041] The second judgment module is used to determine whether the voice information is a wake-up word when the voice information is received in the non-wake-up state.

[0042] A storage module is used to change the current state to a wake-up state and store the wake-up word when the voice information is a wake-up word;

[0043] The first judgment module includes:

[0044] The judgment submodule is used to determine whether the voice information and the wake word stored when changing the current state to the wake state are issued by the same person.

[0045] In one embodiment, the determining submodule is used for:

[0046] The voice information is purified;

[0047] The first voiceprint information of the purified speech information is compared with the second voiceprint information of the wake word stored when the current state is changed to the wake state to determine the similarity between the first voiceprint information and the second voiceprint information.

[0048] When the similarity between the first voiceprint information and the second voiceprint information is greater than a preset similarity, it is determined that the voice command and the wake-up word stored when changing the current state to the wake-up state were issued by the same person.

[0049] In one embodiment, the purification of the voice information includes:

[0050] Extract the audio frames corresponding to the voice information;

[0051] Determine the ratio of signal power to interference power for each audio frame;

[0052] Audio frames whose signal power to interference power ratio is less than a preset ratio are removed. The audio information composed of the remaining audio frames after removing audio frames whose power to interference power ratio is less than the preset ratio is the purified audio information.

[0053] In one embodiment, determining the ratio of signal power to interference power for each audio frame includes:

[0054] When multiple audio signals are detected in each audio frame, the audio signal with the highest power among the multiple audio signals is determined to be the valid signal, and the remaining audio signals are interference signals.

[0055] The power of the valid signal is determined as the signal power of each audio frame, and the power of the interference signal is determined as the interference power of each audio frame.

[0056] Determine the ratio of signal power to interference power for each audio frame.

[0057] In one embodiment, the step of discarding audio frames whose signal power to interference power ratio is less than a preset ratio includes:

[0058] Determine whether the frame length of each audio frame is the same as the speech frame length specified by the speech activity detection model;

[0059] When the frame length of each audio frame is different from the speech frame length specified by the speech activity detection model, a first ratio between the frame length of the audio frame and the speech frame length specified by the speech activity detection model is determined.

[0060] Determine a second ratio between the original index value of the speech activity detection model and the first ratio;

[0061] The multiple integers obtained by rounding down the second ratio are determined as new index values;

[0062] The original index value of the speech activity detection model is replaced with a new index value, wherein after the index value is replaced, the speech activity detection model reads audio frames according to the new index value, and the new index value corresponds to the position of the audio frame;

[0063] Each audio frame, and the ratio of signal power to interference power of each audio frame, are input into the speech activity detection model after the replacement index value. The speech activity detection model is used to purify audio frames by eliminating audio frames whose ratio of signal power to interference power is less than a preset ratio.

[0064] The purified audio frame output by the speech activity detection model after obtaining the replaced index value.

[0065] In one embodiment, the execution module is further configured to:

[0066] The voice command is refused to be executed if it is not issued by the same person as the wake word stored when the user changes to the wake-up state.

[0067] This application also provides a voice information processing system, including:

[0068] At least one processor; and,

[0069] A memory communicatively connected to the at least one processor; wherein,

[0070] The memory stores instructions that can be executed by the at least one processor to implement the voice information processing method described in any of the above embodiments.

[0071] This application also provides a computer-readable storage medium, which, when the instructions in the storage medium are executed by a processor corresponding to a voice information processing system, enables the voice information processing system to implement the voice information processing method described in any of the above embodiments.

[0072] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0073] The technical solution of this application will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0074] The accompanying drawings are provided to further illustrate the present application and form part of the specification. They are used together with the embodiments of the present application to explain the application and do not constitute a limitation thereof. In the drawings:

[0075] Figure 1 This is a flowchart of a voice information processing method according to an embodiment of this application;

[0076] Figure 2 This is a block diagram of a voice information processing device according to an embodiment of this application;

[0077] Figure 3 This is a hardware structure diagram of a voice information processing system according to an embodiment of this application. Detailed Implementation

[0078] The preferred embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit this application.

[0079] Figure 1 This is a flowchart of a voice information processing method according to an embodiment of this application, such as... Figure 1 As shown, the method can be implemented as follows: S101-S103:

[0080] In step S101, when voice information is received in the wake-up state, it is determined whether the voice information and the wake-up word received when the wake-up state is changed are issued by the same person.

[0081] In step S102, when the voice information is issued by the same person as the wake word received when the state is changed to wake-up, the voice information is identified.

[0082] In step S103, the instruction corresponding to the voice information is executed based on the recognition result of the voice information.

[0083] In this application, when voice information is received in the wake-up state, it is determined whether the voice information and the wake word received when changing to the wake-up state were issued by the same person. Specifically, in this application, when voice information is received in the non-wake-up state, it is determined whether the voice information is a wake word. When the voice information is a wake word, the current state is changed to the wake-up state, and the wake word is stored. Since the wake word is stored, when voice information is received in the wake-up state, the stored wake word can be extracted and compared with the voice information to determine whether the voice information and the wake word received when changing to the wake-up state were issued by the same person. When determining whether the voice information and the wake word received when changing to the wake-up state were issued by the same person, the voice information is first purified, and then the first voiceprint information of the purified voice information and the second voiceprint information of the wake word received when changing to the wake-up state are compared to determine the similarity between the first voiceprint information and the second voiceprint information. When the similarity between the first voiceprint information and the second voiceprint information is greater than a preset similarity, it is determined that the voice command and the wake word stored when changing to the wake-up state were issued by the same person. It should be noted that when receiving voice information in a non-wake-up state, the voice information can be purified before identifying whether the voice information is a wake word, thus improving the recognition effect. When changing the current state to a wake-up state and storing the wake word, the purified wake word can also be stored.

[0084] When refining speech information, the corresponding audio frames can be extracted. When multiple audio signals are detected in each audio frame, the audio signal with the highest power among the multiple audio signals is determined as the effective signal, and the remaining audio signals are interference signals. The power of the effective signal is determined as the signal power of each audio frame, and the power of the interference signal is determined as the interference power of each audio frame. The ratio of the signal power to the interference power of each audio frame is determined. Each audio frame, and the ratio of the signal power to the interference power of each audio frame, are input into the speech activity detection model. The speech information consisting of the remaining audio frames after removing audio frames whose power-to-interference power ratio is less than a preset ratio is the refined speech information. The speech activity detection model is used to remove audio frames whose power-to-interference power ratio is less than a preset ratio. The speech activity detection model can be a VAD (Voice Activity Detection) model. In fact, the speech activity detection model can also be implemented by converting audio frames whose signal power to interference power ratio is less than a preset ratio into 0 vectors, audio frames whose signal power to interference power ratio is greater than a preset ratio into 1 vectors, and then discarding the 0 vectors.

[0085] It should be noted that when the speech activity detection model is the VAD model, since the frame shift of the VAD model is 10ms, if the current frame length of the received speech information is 16ms, an alignment operation is required. During alignment, the corresponding alignment rules can be injected into the VAD model to achieve frame alignment.

[0086] When the voice information is issued by the same person as the wake word received when the state is changed to wake-up, the voice information is identified; and the instruction corresponding to the voice information is executed according to the identification result.

[0087] Specifically, in voice interaction scenarios, if multiple users are conversing nearby, when a user wants to interact via voice, they wake up the voice interaction device using a wake word. The device may then misinterpret the conversations of other users as commands, affecting the voice interaction effect. Therefore, in this embodiment, when the recognition result indicates that the voice information is a voice command and the current state is awake, it determines whether the voice command and the wake word received when changing to the awake state were issued by the same person. If the voice command and the wake word stored when changing to the awake state were issued by the same person, the voice command is executed; if the voice command and the wake word stored when changing to the awake state were not issued by the same person, the voice command is rejected. In this way, upon receiving a voice command, it can be determined whether the voice command was issued by the user who issued the wake word. If it was, the voice command is executed; otherwise, it is rejected. In other words, after the voice interaction device is awakened, only the voice command issued by the user who issued the wake word is executed, reducing the impact of other users' voices on the voice interaction and improving the voice interaction effect.

[0088] The beneficial effects of this application are as follows: when receiving voice information, it determines whether the voice information and the wake word received when changing to the wake state are issued by the same person before recognizing the voice information and executing the instruction corresponding to the voice message based on the recognition result. In other words, in the wake state, only the voice instructions issued by the user who issued the wake word are executed, and the voice instructions issued by other users are not executed, which reduces the probability of misrecognition and improves the voice interaction effect.

[0089] In one embodiment, the method may also be implemented as the following steps A1-A2:

[0090] In step A1, when voice information is received in a non-wake-up state, it is determined whether the voice information is a wake-up word;

[0091] In step A2, when the voice information is a wake word, the current state is changed to a wake state, and the wake word is stored;

[0092] In step S101 above, determining whether the voice information and the wake-up word received when changing to the wake-up state are issued by the same person can be implemented as the following step A3:

[0093] In step A3, it is determined whether the voice information and the wake word stored when changing the current state to the wake state were spoken by the same person.

[0094] In one embodiment, step S101 above can be implemented as the following steps B1-B3:

[0095] In step B1, the voice information is purified;

[0096] In step B2, the first voiceprint information of the purified speech information and the second voiceprint information of the wake-up word received when the state is changed to wake-up are compared to determine the similarity between the first voiceprint information and the second voiceprint information.

[0097] In step B3, when the similarity between the first voiceprint information and the second voiceprint information is greater than a preset similarity, it is determined that the voice command and the wake-up word stored when changing to the wake-up state were issued by the same person.

[0098] In one embodiment, step B1 above can be implemented as the following steps C1-C3:

[0099] In step C1, the audio frames corresponding to the voice information are extracted;

[0100] In step C2, the ratio of signal power to interference power for each audio frame is determined;

[0101] In step C3, audio frames whose signal power to interference power ratio is less than a preset ratio are removed. The remaining audio frames after removing audio frames whose power to interference power ratio is less than the preset ratio constitute the purified audio information.

[0102] In an in-vehicle scenario, the voice information can be the sound received by the in-vehicle microphone. After the in-vehicle microphone receives the sound, the front end will use existing noise reduction methods to reduce the noise before transmitting it to the execution subject of this application. That is, the voice information received by the execution subject of this application is already the voice information after noise reduction by the front end. The front end sends voice information in the form of audio frames. Therefore, the audio frames corresponding to the voice information can be extracted. Typically, the frame shift of the front end is usually 16ms. Therefore, the frame length of the audio frame sent by the front end is usually 16ms. That is to say, a 1-second voice information is approximately 60 audio frames. The ratio of signal power to interference power of each audio frame is determined. Audio frames in each audio frame whose ratio of signal power to interference power is less than a preset ratio are removed. The voice information composed of the remaining audio frames after removing audio frames whose ratio of power to interference power is less than the preset ratio is the purified voice information.

[0103] In one embodiment, step C1 above can be implemented as follows:

[0104] In step D1, when multiple audio signals are detected in each audio frame, the audio signal with the highest power among the multiple audio signals is determined to be the valid signal, and the remaining audio signals are interference signals.

[0105] In step D2, the power of the effective signal is determined as the signal power of each audio frame, and the power of the interference signal is determined as the interference power of each audio frame.

[0106] In step D3, the ratio of signal power to interference power for each audio frame is determined.

[0107] Specifically, if multiple audio signals are detected in an audio frame, the audio signal with the highest power among the multiple audio signals is determined as the valid signal, and the remaining audio signals are interference signals; the power of the valid signal is determined as the signal power of each audio frame, and the power of the interference signal is determined as the interference power of each audio frame; the ratio of the signal power to the interference power of each audio frame is then determined. For example, the ratio of the signal power to the interference power of each audio frame can be determined using the following formula:

[0108]

[0109] Among them, P s For signal power, P I For interference power, P s and P I It can be estimated using classic signal processing algorithms. The specific principle is to determine the effective signal from among multiple audio signals based on the highest power. SIR is the ratio of signal power to interference power.

[0110] In one embodiment, step C3 above can be implemented as the following steps E1-E7:

[0111] In step E1, it is determined whether the frame length of each audio frame is the same as the speech frame length specified by the speech activity detection model;

[0112] In step E2, when the frame length of each audio frame is different from the speech frame length specified by the speech activity detection model, a first ratio between the frame length of the audio frame and the speech frame length specified by the speech activity detection model is determined.

[0113] In step E3, a second ratio is determined between the original index value of the speech activity detection model and the first ratio;

[0114] In step E4, the multiple integers obtained after rounding down the second ratio are determined as new index values;

[0115] In step E5, the original index value of the speech activity detection model is replaced with a new index value. After the index value is replaced, the speech activity detection model reads audio frames according to the new index value, and the new index value corresponds to the position of the audio frame.

[0116] In step E6, each audio frame and the ratio of signal power to interference power of each audio frame are input into the speech activity detection model after the replacement index value. The speech activity detection model is used to purify audio frames by eliminating audio frames whose ratio of signal power to interference power is less than a preset ratio.

[0117] In step E7, the purified audio frame output by the speech activity detection model after the replacement index value is obtained.

[0118] In this embodiment, during purification, each audio frame, and the ratio of signal power to interference power for each audio frame, can be input into a speech activity detection model to purify the speech information. The speech activity detection model is used to remove audio frames whose signal power to interference power ratio is less than a preset ratio. This speech activity detection model can be a VAD (Voice Activity Detection) model. The speech activity detection model can purify the speech information by converting audio frames whose signal power to interference power ratio is less than a preset ratio into 0 vectors, and audio frames whose signal power to interference power ratio is greater than a preset ratio into 1 vectors, and then removing the 0 vectors.

[0119] It should be noted that the frame length of the audio frame corresponding to the speech activity detection model and the speech information received from the front end may differ. Therefore, alignment is required before inputting the audio frame and the ratio of audio frame signal power to interference power into the speech activity detection model. Specifically, it is determined whether the frame length of each audio frame is the same as the speech frame length specified by the speech activity detection model. When the frame length of each audio frame is different from the speech frame length specified by the speech activity detection model, a first ratio between the frame length of the audio frame and the speech frame length specified by the speech activity detection model is determined. A second ratio between the original index value of the speech activity detection model and the first ratio is determined. The integers obtained by rounding down the second ratio are determined as new index values. The original index value of the speech activity detection model is replaced with the new index value. After replacing the index value, the speech activity detection model reads the audio frame according to the new index value, and the new index value corresponds to the position of the audio frame.

[0120] For example, when the speech activity detection model is the VAD model, since the frame shift of the VAD model is 10ms (that is, the speech frame length specified by the speech activity detection model is 10ms, and the speech activity detection model reads audio frames in 10ms increments), if the current frame length of the received speech information is 16ms, an alignment operation is required. During alignment, corresponding alignment rules can be injected into the VAD model. The alignment rule can be the index value i in the VAD model divided by 1.6 and rounded down. This index value is used to guide the VAD model to read the i-th audio frame. Under this alignment rule, when the index value is 1, dividing by 1.6 and rounding down results in the index becoming 0; when the index value is 2, dividing by 1.6 and rounding down results in the index becoming 1; when the index value is 3, dividing by 1.6 and rounding down results in the index becoming 1; when the index value is 4, dividing by 1.6 and rounding down results in the index becoming 2, and so on. Through this alignment rule, it can be seen that when the original index value is 1, no audio is read; when the original index value is 2 or 3, the first audio frame will be read; when the original index is 4, the second audio frame will be read; and when the original index is 96-100, the 60th audio frame will be read.

[0121] After alignment, each audio frame, along with the ratio of signal power to interference power for each audio frame, is input into the speech activity detection model after the replacement index value. The speech activity detection model is used to purify audio frames by eliminating audio frames whose signal power to interference power ratio is less than a preset ratio. The purified audio frames output by the speech activity detection model after the replacement index value are then obtained.

[0122] In one embodiment, the method may also be implemented as follows:

[0123] The voice command is refused to be executed if it is not issued by the same person as the wake word stored when the user changes to the wake-up state.

[0124] In this embodiment, if the voice command is not issued by the same person as the wake word stored when the device is changed to the wake-up state, the voice command will not be executed. In other words, after the voice interaction device is woken up, only the voice command issued by the user who issued the wake word is executed, reducing the impact of other users' voices on the voice interaction and improving the voice interaction effect.

[0125] Figure 2 A block diagram of a voice information processing device, such as Figure 2 As shown, the device may include the following modules:

[0126] The first judgment module 201 is used to determine whether the voice information received in the wake-up state is issued by the same person as the wake-up word received when the wake-up state is changed.

[0127] The recognition module 202 is used to recognize the voice information when the voice information is issued by the same person as the wake word received when the state is changed to wake-up.

[0128] The execution module 203 is used to execute the instruction corresponding to the voice information based on the recognition result of the voice information.

[0129] In one embodiment, the apparatus further includes:

[0130] The second judgment module is used to determine whether the voice information is a wake-up word when the voice information is received in the non-wake-up state.

[0131] A storage module is used to change the current state to a wake-up state and store the wake-up word when the voice information is a wake-up word;

[0132] The first judgment module includes:

[0133] The judgment submodule is used to determine whether the voice information and the wake word stored when changing the current state to the wake state are issued by the same person.

[0134] In one embodiment, the determining submodule is used for:

[0135] The voice information is purified;

[0136] The first voiceprint information of the purified speech information is compared with the second voiceprint information of the wake word stored when the current state is changed to the wake state to determine the similarity between the first voiceprint information and the second voiceprint information.

[0137] When the similarity between the first voiceprint information and the second voiceprint information is greater than a preset similarity, it is determined that the voice command and the wake-up word stored when changing the current state to the wake-up state were issued by the same person.

[0138] In one embodiment, the purification of the voice information includes:

[0139] Extract the audio frames corresponding to the voice information;

[0140] Determine the ratio of signal power to interference power for each audio frame;

[0141] Audio frames whose signal power to interference power ratio is less than a preset ratio are removed. The audio information composed of the remaining audio frames after removing audio frames whose power to interference power ratio is less than the preset ratio is the purified audio information.

[0142] In one embodiment, determining the ratio of signal power to interference power for each audio frame includes:

[0143] When multiple audio signals are detected in each audio frame, the audio signal with the highest power among the multiple audio signals is determined to be the valid signal, and the remaining audio signals are interference signals.

[0144] The power of the valid signal is determined as the signal power of each audio frame, and the power of the interference signal is determined as the interference power of each audio frame.

[0145] Determine the ratio of signal power to interference power for each audio frame.

[0146] In one embodiment, the step of discarding audio frames whose signal power to interference power ratio is less than a preset ratio includes:

[0147] Determine whether the frame length of each audio frame is the same as the speech frame length specified by the speech activity detection model;

[0148] When the frame length of each audio frame is different from the speech frame length specified by the speech activity detection model, a first ratio between the frame length of the audio frame and the speech frame length specified by the speech activity detection model is determined.

[0149] Determine a second ratio between the original index value of the speech activity detection model and the first ratio;

[0150] The multiple integers obtained by rounding down the second ratio are determined as new index values;

[0151] The original index value of the speech activity detection model is replaced with a new index value, wherein after the index value is replaced, the speech activity detection model reads audio frames according to the new index value, and the new index value corresponds to the position of the audio frame;

[0152] Each audio frame, and the ratio of signal power to interference power of each audio frame, are input into the speech activity detection model after the replacement index value. The speech activity detection model is used to purify audio frames by eliminating audio frames whose ratio of signal power to interference power is less than a preset ratio.

[0153] The purified audio frame output by the speech activity detection model after obtaining the replaced index value.

[0154] In one embodiment, the execution module is further configured to:

[0155] The voice command is refused to be executed if it is not issued by the same person as the wake word stored when the user changes to the wake-up state.

[0156] Figure 3 This is a schematic diagram of the hardware structure of a voice information processing system according to an embodiment of this application, as shown below. Figure 3 As shown, the voice information processing system includes:

[0157] At least one processor 320; and,

[0158] Memory 304 communicatively connected to the at least one processor 320; wherein,

[0159] The memory 304 stores instructions that can be executed by the at least one processor 320 to implement the voice information processing method described in any of the above embodiments.

[0160] Reference Figure 3 The voice information processing system 300 may include one or more of the following components: processing component 302, memory 304, power supply component 306, multimedia component 308, audio component 310, input / output (I / O) interface 312, sensor component 314, and communication component 316.

[0161] Processing component 302 typically controls the overall operation of voice information processing system 300. Processing component 302 may include one or more processors 320 to execute instructions to complete all or part of the steps of the methods described above. Furthermore, processing component 302 may include one or more modules to facilitate interaction between processing component 302 and other components. For example, processing component 302 may include a multimedia module to facilitate interaction between multimedia component 308 and processing component 302.

[0162] Memory 304 is configured to store various types of data to support the operation of voice information processing system 300. Examples of this data include instructions for any application or method operating on voice information processing system 300, such as text, images, video, etc. Memory 304 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 storage, flash memory, magnetic disk, or optical disk.

[0163] The power supply assembly 306 provides power to the various components of the voice information processing system 300. The power supply assembly 306 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the vehicle control system 300.

[0164] Multimedia component 308 includes a screen that provides an output interface between voice information processing system 300 and user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of touch or swipe actions but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 308 may also include a front-facing camera and / or a rear-facing camera. When voice information processing system 300 is in an operating mode, such as shooting mode or video mode, the front-facing camera and / or rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0165] Audio component 310 is configured to output and / or input audio signals. For example, audio component 310 includes a microphone (MIC) configured to receive external audio signals when the voice information processing system 300 is in an operating mode, such as alarm mode, recording mode, voice recognition mode, and voice output mode. The received audio signals may be further stored in memory 304 or transmitted via communication component 316. In some embodiments, audio component 310 also includes a speaker for outputting audio signals.

[0166] I / O interface 312 provides an interface between processing component 302 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0167] Sensor assembly 314 includes one or more sensors for providing status assessments of various aspects of the voice information processing system 300. For example, sensor assembly 314 may include a sound sensor. Additionally, sensor assembly 314 can detect the on / off state of the voice information processing system 300, the relative positioning of components (e.g., the display and keypad of the voice information processing system 300), and the operating state of the voice information processing system 300 or a component thereof (e.g., the operating state of the air distribution plate, structural state, discharge scraper, etc.), the orientation or acceleration / deceleration of the voice information processing system 300, and temperature changes of the voice information processing system 300. Sensor assembly 314 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 314 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 314 may also include an accelerometer, a gyroscope, a magnetometer, a pressure sensor, a material buildup thickness sensor, or a temperature sensor.

[0168] Communication component 316 is configured to enable voice information processing system 300 to provide wired or wireless communication capabilities with other devices and cloud platforms. Voice information processing system 300 can access wireless networks based on communication standards, such as WiFi, 2G, or 3G, or combinations thereof. In one exemplary embodiment, communication component 316 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 316 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0169] In an exemplary embodiment, the voice information processing system 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the voice information processing method described in any of the above embodiments.

[0170] This application also provides a computer-readable storage medium, which, when the instructions in the storage medium are executed by a processor corresponding to a voice information processing system, enables the voice information processing system to implement the voice information processing method described in any of the above embodiments.

[0171] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0172] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0173] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0174] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0175] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for processing voice information, characterized in that, include: When a voice message is received in the wake-up state, it is determined whether the voice message and the wake-up word received when the wake-up state is changed are issued by the same person. When the voice information is issued by the same person as the wake word received when the state is changed to wake-up, the voice information is identified. Execute the instruction corresponding to the voice information based on the voice information recognition result; The step of determining whether the voice information and the wake-up word stored when changing the current state to the wake-up state are from the same person includes purifying the voice information; the step of removing audio frames whose signal power to interference power ratio is less than a preset ratio in the voice information purification step is specifically implemented in the following way: Determine whether the frame length of each audio frame is the same as the speech frame length specified by the speech activity detection model; When the frame length of each audio frame is different from the speech frame length specified by the speech activity detection model, a first ratio between the frame length of the audio frame and the speech frame length specified by the speech activity detection model is determined. Determine a second ratio between the original index value of the speech activity detection model and the first ratio; The multiple integers obtained by rounding down the second ratio are determined as new index values; The original index value of the speech activity detection model is replaced with a new index value, wherein after the index value is replaced, the speech activity detection model reads audio frames according to the new index value, and the new index value corresponds to the position of the audio frame; Each audio frame, and the ratio of signal power to interference power of each audio frame, are input into the speech activity detection model after the replacement index value. The speech activity detection model is used to purify audio frames by eliminating audio frames whose ratio of signal power to interference power is less than a preset ratio. The purified audio frame output by the speech activity detection model after obtaining the replaced index value.

2. The method as described in claim 1, characterized in that, The method further includes: When a voice message is received in a non-wake-up state, it is determined whether the voice message is a wake-up word; When the voice information is a wake word, the current state is changed to a wake state, and the wake word is stored; The step of determining whether the voice information and the wake-up word received when changing to the wake-up state are issued by the same person includes: Determine whether the voice information and the wake word stored when changing the current state to the wake state were spoken by the same person.

3. The method as described in claim 1, characterized in that, The step of determining whether the voice information and the wake word stored when changing the current state to the wake state were spoken by the same person includes: The voice information is purified; The first voiceprint information of the purified speech information is compared with the second voiceprint information of the wake word stored when the current state is changed to the wake state to determine the similarity between the first voiceprint information and the second voiceprint information. When the similarity between the first voiceprint information and the second voiceprint information is greater than a preset similarity, it is determined that the instruction corresponding to the voice information and the wake-up word stored when changing the current state to the wake-up state were issued by the same person.

4. The method as described in claim 3, characterized in that, The purification of the voice information includes: Extract the audio frames corresponding to the voice information; Determine the ratio of signal power to interference power for each audio frame; Audio frames whose signal power to interference power ratio is less than a preset ratio are removed. The audio information composed of the remaining audio frames after removing audio frames whose power to interference power ratio is less than the preset ratio is the purified audio information.

5. The method as described in claim 4, characterized in that, Determining the ratio of signal power to interference power for each audio frame includes: When multiple audio signals are detected in each audio frame, the audio signal with the highest power among the multiple audio signals is determined to be the valid signal, and the remaining audio signals are interference signals. The power of the valid signal is determined as the signal power of each audio frame, and the power of the interference signal is determined as the interference power of each audio frame. Determine the ratio of signal power to interference power for each audio frame.

6. The method as described in claim 1, characterized in that, The method further includes: If the instruction corresponding to the voice information is not issued by the same person as the wake word stored when the state is changed to wake-up, the instruction corresponding to the voice information will not be executed.

7. A voice information processing apparatus for implementing the voice information processing method as described in any one of claims 1-6, characterized in that, include: The first judgment module is used to determine whether the voice information received in the wake-up state is issued by the same person as the wake-up word received when the wake-up state is changed. The recognition module is used to recognize the voice information when the voice information is issued by the same person as the wake word received when the state is changed to wake-up. The execution module is used to execute the instructions corresponding to the voice information based on the recognition result of the voice information.

8. A voice information processing system, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to implement the voice information processing method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor corresponding to the voice information processing system, the voice information processing system is able to implement the voice information processing method as described in any one of claims 1-6.

Images