Voice wake-up method and device, electronic equipment and storage medium

Abstract

The invention discloses a voice wake-up method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the accent type recognition of target wake-up audio information if the target wake-up audio information is detected, and obtaining a target accent type corresponding to the target wake-up audio information; determining a target accent wake-up model corresponding to the target accent type based on a corresponding relationship between the accent type and an accent wake-up model; and processing the target wake-up audio information based on the target accent wake-up model to execute a wake-up operation. According to the scheme, the problem that awakening is difficult due to nonstandard accent can be effectively solved, the awakening effect is effectively improved, and the awakening experience is optimized.

Description

Technical Field

[0001] The present invention belongs to the field of voice processing, and in particular relates to a voice wake-up method, device, electronic equipment and storage medium. Background Art

[0002] With the continuous development of science and technology, voice control is applied to more and more scenarios. When controlling devices or applications through voice, users are usually required to say a wake-up word to wake up the controlled object. In related technologies, when the user's pronunciation is inaccurate or the user has an accent, the wake-up effect will be greatly reduced, making it difficult to wake up the controlled object, or even impossible to wake up, affecting the user experience. Summary of the invention

[0003] In view of the above technical problems in the related art, the embodiments of the present invention provide a voice wake-up method, device, electronic device and storage medium to improve the wake-up effect and optimize the wake-up experience.

[0004] In a first aspect, an embodiment of the present invention provides a voice wake-up method, the method comprising:

[0005] If target wake-up audio information is detected, performing accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information;

[0006] Determining a target accent awakening model corresponding to the target accent type based on a correspondence between the accent type and the accent awakening model;

[0007] The target wake-up audio information is processed based on the target accent wake-up model to perform a wake-up operation.

[0008] In some embodiments, before detecting the target wake-up audio information and performing accent type recognition on the target wake-up audio information, the method further includes:

[0009] receiving first audio information;

[0010] The first audio information is matched with the reference wake-up audio information. If the match is successful, it is determined that the target wake-up audio information is detected, and the reference wake-up audio information is the audio information corresponding to the wake-up word.

[0011] In some implementations, the performing accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information includes:

[0012] Based on a preset accent classification model, the target wake-up audio is identified to obtain the target accent type corresponding to the target wake-up audio, wherein the target accent type is a standard accent type, a dialect accent type, or a pronunciation deviation accent type.

[0013] In some implementations, each accent awakening model in the corresponding relationship is obtained by the following steps:

[0014] For each accent wake-up model, an initial wake-up model is constructed and a training sample set is obtained, wherein each sample in the training sample set is audio information under a corresponding accent type;

[0015] Based on the training sample set under each accent type, the corresponding initial awakening model is trained to obtain a trained accent awakening model.

[0016] In some implementations, obtaining a training sample set includes:

[0017] Acquiring second audio information;

[0018] Matching the second audio information with each piece of audio information in a preset accent library, wherein the preset accent library includes a plurality of preset audio information with accent tags, each accent tag corresponding to an accent type;

[0019] If the matching result shows that the similarity between the second audio information and the N preset audio information in the preset accent library is greater than or equal to a threshold, the accent tag of the preset audio information corresponding to the maximum similarity is used as the accent tag of the second audio information, where N is a positive integer;

[0020] Based on the accent label of the second audio information, the second audio information is divided into a training sample set under a corresponding accent type.

[0021] In some embodiments, after matching the second audio information with each piece of audio information in a preset accent library, the method further includes:

[0022] If the matching result indicates that the similarity between the second audio information and each piece of preset audio information in the preset accent library is less than the threshold, outputting update confirmation information for the preset accent library, wherein the update confirmation information is used to determine whether the accent tag of the second audio information is a newly added accent tag;

[0023] receiving feedback information in response to the update confirmation information;

[0024] If the feedback information confirms that the accent tag of the second audio information is a newly added accent tag, the accent tag corresponding to the second audio information is obtained, and the second audio information and the corresponding accent tag are added to the preset accent library.

[0025] In some implementations, after performing accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information, the method further includes:

[0026] A target speech recognition model corresponding to the target accent type is determined to process the received voice instruction through the target speech recognition model.

[0027] In a second aspect, an embodiment of the present invention provides a voice wake-up device, including:

[0028] An accent recognition module, configured to, if target wake-up audio information is detected, perform accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information;

[0029] A model determination module, configured to determine a target accent awakening model corresponding to the target accent type based on a correspondence between the accent type and the accent awakening model;

[0030] A processing module is used to process the target wake-up audio information based on the target accent wake-up model to perform a wake-up operation.

[0031] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any embodiment of the first aspect when executing the program.

[0032] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any embodiment of the first aspect.

[0033] One or more technical solutions provided by the embodiments of the present invention achieve at least the following technical effects or advantages:

[0034] The voice wake-up method provided in the embodiment of this specification, when detecting the target wake-up audio information, performs accent type recognition on the target wake-up audio information to obtain the target accent type corresponding to the target wake-up audio information; based on the correspondence between the accent type and the accent wake-up model, determines the target accent wake-up model corresponding to the target accent type; and processes the target wake-up audio information based on the target accent wake-up model to perform the wake-up operation. In the above scheme, by identifying the accent type of the target wake-up audio information and using different wake-up models for different accent types to process the target wake-up audio information, the problem of difficulty in wake-up caused by non-standard accents can be effectively solved, the wake-up effect is effectively improved, and the wake-up experience is optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 A flow chart of a voice wake-up method provided in an embodiment of this specification;

[0037] Figure 2 A flowchart of another voice wake-up method provided in an embodiment of this specification;

[0038] Figure 3 A schematic diagram of a voice wake-up device provided in an embodiment of this specification;

[0039] Figure 4 A schematic diagram of an electronic device provided in an embodiment of this specification. DETAILED DESCRIPTION

[0040] The embodiments of the present specification provide a voice wake-up method, device, electronic device and storage medium, the method comprising: if target wake-up audio information is detected, performing accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information; based on the correspondence between the accent type and the accent wake-up model, determining a target accent wake-up model corresponding to the target accent type; and processing the target wake-up audio information based on the target accent wake-up model to perform a wake-up operation.

[0041] The solution in the embodiments of this specification can effectively solve the problem of difficulty in awakening due to non-standard accents by identifying the accent type of the target wake-up audio information and using different wake-up models to process the target wake-up audio information according to different accent types, thereby effectively improving the wake-up effect and optimizing the wake-up experience.

[0042] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0044] like Figure 1 FIG. 1 is a flowchart of a voice wake-up method provided in an embodiment of the present specification, and the method includes the following steps:

[0045] Step S101: if target wake-up audio information is detected, perform accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information;

[0046] Step S102: determining a target accent awakening model corresponding to the target accent type based on the correspondence between the accent type and the accent awakening model;

[0047] Step S103: Processing the target wake-up audio information based on the target accent wake-up model to perform a wake-up operation.

[0048] The method provided in the embodiments of this specification can be used in electronic devices, for example, in mobile phones, tablet computers, household appliances, etc. that provide voice control, and is not limited here.

[0049] In step S101, the electronic device may be provided with an audio acquisition module, and the audio information output by the user may be collected in real time through the audio acquisition module. In some embodiments, the target wake-up audio information may be any collected audio information. In some embodiments, in order to save computing resources during the wake-up process, the target wake-up audio information may be audio information that has been preliminarily screened and is similar to the wake-up word.

[0050] During the specific implementation process, the target wake-up audio information can be obtained in the following manner: receiving the first audio information; matching the first audio information with the reference wake-up audio information. If the match is successful, it is determined that the target wake-up audio information is detected, and the reference wake-up audio information is the audio information corresponding to the wake-up word.

[0051] Specifically, in a wake-up event, the user needs to say the wake-up word to trigger the electronic device to perform the wake-up operation. The wake-up word can be preset, and the wake-up word can be different for different awakened objects. For example, when waking up the air conditioner, the wake-up word is A, and when waking up the smart speaker, the wake-up word is B.

[0052] The electronic device can continuously collect audio information output by the user through the audio collection module, wherein the first audio information can be any audio information collected by the electronic device. In order to save computing resources for the wake-up event, the first audio information can be preliminarily judged to determine whether the first audio information contains the wake-up word or information similar to the wake-up word. If it does, the subsequent accent type recognition and wake-up processing are further performed. If the first audio information does not contain the wake-up word, it is considered that the first audio information is not the audio information for performing the wake-up operation, and the first audio information can be directly discarded without any processing.

[0053] When determining whether the first audio information contains the wake-up word, it can be achieved by matching the first audio information with the reference wake-up audio information. The reference wake-up audio information can be audio information corresponding to the wake-up word. For example, the reference wake-up audio information can be audio information when the wake-up word is read out with a standard accent. For another example, the audio information of the wake-up word is collected when various accents are used to read out the wake-up word, and the audio features of the audio information corresponding to each accent are extracted, and the reference wake-up audio information is generated based on the extracted audio features.

[0054] Matching the first audio information with the reference wake-up audio information may be to calculate the similarity between the first audio information and the reference wake-up audio information, and when the similarity is greater than a preset similarity, the match is considered successful. Matching the first audio information with the reference wake-up audio information may also be to compare the amplitude of the first audio information with the amplitude of the reference wake-up audio information, and if the amplitude error between the two meets a preset range, the match is considered successful. Of course, other methods may also be used to determine whether the first audio information matches the reference wake-up audio information, which is not limited here.

[0055] When the first audio information successfully matches the reference wake-up audio information, it indicates that the first audio information may correspond to the wake-up word. At this time, the first audio information can be used as the target wake-up audio information for subsequent processing.

[0056] In order to avoid the poor wake-up effect when users with inaccurate pronunciation or accents perform the wake-up operation, in an embodiment of the present specification, after detecting the target wake-up audio information, the target wake-up audio information is subjected to accent type recognition to obtain the target accent type. Among them, the target accent type may include but is not limited to the standard accent type, the dialect accent type, and the pronunciation deviation accent type. Among them, the standard accent type may be a universal standard pronunciation, the dialect accent type may include various dialect accents, and the pronunciation deviation accent may be a non-standard pronunciation, such as the non-standard pronunciation of foreigners, etc. It should be noted that the pronunciation deviation accent is neither a standard accent nor a dialect.

[0057] In some embodiments, accent type recognition may be achieved in the following manner: based on a preset accent classification model, the target wake-up audio is recognized to obtain the target accent type corresponding to the target wake-up audio.

[0058] The preset accent classification model may be a pre-trained model, wherein the preset accent classification model may be a convolutional neural network model, a recurrent neural network model, etc., which is not limited here. In the specific training process of the model, the training samples may include audio information under various accent types. For example, if the accent types output by the preset accent classification model include standard accent types, dialect accent types, and pronunciation deviation accent types, then the training samples need to cover the audio information under the standard accent types, dialect accent types, and pronunciation deviation accent types. Furthermore, model training is performed based on the training samples, and the trained model is used as the preset accent classification model.

[0059] The target wake-up audio information is input into a preset accent classification model to obtain the target accent type of the target wake-up audio information.

[0060] Step S102, after determining the target accent type, the target accent wake-up model can be determined by querying the correspondence between the accent type and the accent wake-up model. The correspondence between the accent type and the accent wake-up model can be pre-set. For example, the standard accent type corresponds to the standard accent wake-up model, the dialect accent type corresponds to the dialect accent wake-up model, and the pronunciation deviation type accent type corresponds to the pronunciation deviation type accent wake-up model.

[0061] In the embodiments of the present specification, corresponding accent awakening models can be pre-constructed for different accent types, and the construction process for each accent awakening model can be implemented through the following steps: for each accent awakening model, an initial awakening model is constructed and a training sample set is obtained, wherein each sample in the training sample set is audio information under the corresponding accent type; based on the training sample set under each accent type, the corresponding initial awakening model is trained to obtain a trained accent awakening model.

[0062] It should be noted that the initial awakening model corresponding to each accent awakening model can select the specific type of model according to actual needs, such as convolutional neural network model, recurrent neural network model, etc., which is not limited here. Still taking the accent type including standard accent type, dialect accent type and pronunciation deviation type as an example, then, it is necessary to construct an initial awakening model of the standard accent awakening model for processing the audio information of the standard accent type, an initial awakening model of the dialect accent awakening model for processing the audio of the dialect accent type, and an initial awakening model of the pronunciation deviation type accent awakening model for processing the audio of the pronunciation deviation type accent type.

[0063] When training each accent wake-up model, it is necessary to obtain a set of training samples under the corresponding accent type. In some embodiments, the training sample set can be manually screened, for example, by collecting audio information of wake-up words read by users with different accents, and marking the accent type corresponding to the audio information, and adding audio information with the same label to the same set to obtain training sample sets under different accent types.

[0064] In order to achieve automatic labeling of training samples, in an embodiment of the present specification, a training sample set can also be obtained in the following manner: obtain second audio information; match the second audio information with each audio information in a preset accent library, wherein the preset accent library includes multiple preset audio information with accent tags, and each accent tag corresponds to an accent type; if the matching result shows that the similarity between the second audio information and N preset audio information in the preset accent library is greater than or equal to a threshold, then the accent tag of the preset audio information corresponding to the maximum similarity is used as the accent tag of the second audio information, where N is a positive integer; based on the accent tag of the second audio information, the second audio information is divided into a training sample set under the corresponding accent type.

[0065] Specifically, a preset accent library can be pre-constructed, wherein the preset accent library contains preset audio information for each accent type. For example, the preset accent library contains preset audio information for reading the wake-up word with a standard accent, and the accent label of the preset audio information is the standard accent; the preset accent library contains preset audio information for reading the wake-up word with a Cantonese accent, and the accent label of the preset audio information is the Cantonese accent; the preset accent library contains preset audio information for reading the wake-up word with a Sichuan accent, and the accent label of the preset audio information is the Sichuan accent. Of course, the preset accent library can also contain preset audio information corresponding to the deviant accent, which will not be listed here one by one. In the above example, the standard accent corresponds to the standard accent type, and the Cantonese accent and Sichuan accent correspond to the dialect accent type. It should be noted that in the preset accent library, there can be at least one preset audio information under each accent label.

[0066] The second audio information can be the audio information output by any user that has been collected. The second audio information is matched with each preset audio information in the preset accent library. For example, the second audio information is similar to each preset audio information. If the similarity between the second audio information and N preset audio information is greater than or equal to the threshold, it means that the accent corresponding to the second audio information is similar to the accent of these N preset audio information. Then, the accent label of the audio information with the greatest similarity can be used as the accent label of the second audio information. It should be noted that the threshold can be selected according to actual needs. For example, the threshold can be 90%, 95%, etc., which is not limited here.

[0067] After the accent label of the second audio information is obtained, since each accent label corresponds to an accent type, the second audio information can be used as a training sample under the accent type.

[0068] In addition, if the similarity between the second audio information and each preset audio is less than a threshold, it means that the second audio information is not similar to the accents contained in the preset accent library, so the second audio information may be other dialects or a deviant accent. At this time, the accent type of the second audio information can be determined by the following steps: if the matching result shows that the similarity between the second audio information and each preset audio information in the preset accent library is less than the threshold, output update confirmation information for the preset accent library, wherein the update confirmation information is used to determine whether the accent tag of the second audio information is a newly added accent tag; receive feedback information in response to the update confirmation information; if the feedback information is to confirm that the accent tag of the second audio information is a newly added accent tag, obtain the accent tag corresponding to the second audio information, and add the second audio information and the corresponding accent tag to the preset accent library.

[0069] Specifically, when the similarity between the second audio information and each preset audio information is less than a threshold, the user may be asked whether the second audio information is a newly added accent, wherein the update confirmation information for the preset accent library may be output through an interactive interface, audio interaction, or push to the user's terminal device. Taking the interactive interface as an example, a dialog box may be displayed on the interactive interface to ask the user whether the accent tag of the second audio information is a newly added accent tag, and two options of "yes" and "no" may be provided. When the user selects "yes", the user may be further reminded to enter the accent tag of the second audio information, and the second audio information and the corresponding accent tag may be added as newly added preset audio information and newly added accent tags to the preset accent library. If the user's choice is "no", the second audio information may be treated as a deviant accent type, and the second audio information may be labeled with a deviant accent.

[0070] It can be seen that by automatically matching the preset audio information in the preset accent library, the automatic labeling of training samples can be achieved, which greatly reduces the dependence on manual labeling. After obtaining the training sample set under each accent type, the respective initial awakening models can be trained to obtain the trained accent awakening model, and the corresponding relationship between the accent awakening model and the corresponding accent type can be established.

[0071] In step S103, after the target accent awakening model is determined, the target accent awakening model is used to process the target awakening audio information to complete the awakening.

[0072] In the embodiments of the present specification, the target wake-up audio information can be further processed by the target accent wake-up model to accurately identify whether the target wake-up audio information contains the wake-up word. If the target wake-up audio information contains the wake-up word, the wake-up operation is performed; otherwise, it is considered that the target wake-up audio information cannot trigger the wake-up operation.

[0073] It should be noted that after the wake-up operation is completed, the user usually uses voice to implement certain functions. Since the user's target accent type has been determined during the wake-up process, the target speech recognition model corresponding to the target accent type can be determined in the subsequent voice control process to process the received voice commands through the target speech recognition model.

[0074] For ease of understanding, the following takes the accent types including the marked accent type, the dialect accent type and the pronunciation deviation accent type as an example to illustrate the voice wake-up process provided in the embodiment of this specification. Please refer to Figure 2 , the process includes:

[0075] Step S201: Acquire audio information output by the user;

[0076] Step S202: Determine whether the audio information matches the reference wake-up audio information;

[0077] If not, execute step S203, if yes, execute step S204;

[0078] Step S203: discarding audio information;

[0079] Step S204: classifying the audio information by accent type;

[0080] If the audio information is of a standard accent type, execute step S205; if the audio information is of a dialect accent type, execute step S206; if the audio information is of a pronunciation deviation accent type, execute step S207;

[0081] Step S205: Processing the audio information through a standard accent wake-up model to output a wake-up event;

[0082] Step S206: Processing the audio information through the dialect accent wake-up model to output a wake-up event;

[0083] Step S207: Process the audio information through the pronunciation deviation accent wake-up model to output a wake-up event.

[0084] In summary, the solution in the embodiments of this specification solves the wake-up problem caused by non-standard accents by identifying the accent of audio information and selecting different wake-up models to perform the wake-up operation, effectively improving the wake-up effect and optimizing the wake-up experience.

[0085] Based on the same inventive concept, an embodiment of the present invention provides a voice wake-up device, referring to Figure 3 As shown, the device comprises:

[0086] The accent recognition module 301 is used to identify the accent type of the target wake-up audio information if the target wake-up audio information is detected, and obtain the target accent type corresponding to the target wake-up audio information;

[0087] A model determination module 302, configured to determine a target accent awakening model corresponding to the target accent type based on a correspondence between the accent type and the accent awakening model;

[0088] The processing module 303 is used to process the target wake-up audio information based on the target accent wake-up model to perform a wake-up operation.

[0089] In some embodiments, the device further comprises:

[0090] A receiving module, used for receiving first audio information;

[0091] A matching module is used to match the first audio information with reference wake-up audio information. If the match is successful, it is determined that the target wake-up audio information is detected, and the reference wake-up audio information is the audio information corresponding to the wake-up word.

[0092] In some embodiments, the accent recognition module 301 is used to:

[0093] Based on a preset accent classification model, the target wake-up audio is identified to obtain the target accent type corresponding to the target wake-up audio, wherein the target accent type is a standard accent type, a dialect accent type, or a pronunciation deviation accent type.

[0094] In some embodiments, the device further comprises a model building module for:

[0095] For each accent wake-up model, an initial wake-up model is constructed and a training sample set is obtained, wherein each sample in the training sample set is audio information under a corresponding accent type;

[0096] Based on the training sample set under each accent type, the corresponding initial awakening model is trained to obtain a trained accent awakening model.

[0097] In some embodiments, the model building module is used to:

[0098] Acquiring second audio information;

[0099] Matching the second audio information with each piece of audio information in a preset accent library, wherein the preset accent library includes a plurality of preset audio information with accent tags, each accent tag corresponding to an accent type;

[0100] If the matching result shows that the similarity between the second audio information and the N preset audio information in the preset accent library is greater than or equal to a threshold, the accent tag of the preset audio information corresponding to the maximum similarity is used as the accent tag of the second audio information, where N is a positive integer;

[0101] Based on the accent label of the second audio information, the second audio information is divided into a training sample set under a corresponding accent type.

[0102] In some embodiments, the apparatus further comprises an updating module, configured to:

[0103] If the matching result indicates that the similarity between the second audio information and each piece of preset audio information in the preset accent library is less than the threshold, outputting update confirmation information for the preset accent library, wherein the update confirmation information is used to determine whether the accent tag of the second audio information is a newly added accent tag;

[0104] receiving feedback information in response to the update confirmation information;

[0105] If the feedback information confirms that the accent tag of the second audio information is a newly added accent tag, the accent tag corresponding to the second audio information is obtained, and the second audio information and the corresponding accent tag are added to the preset accent library.

[0106] In some embodiments, the device further comprises:

[0107] The voice instruction processing module is used to determine a target voice recognition model corresponding to the target accent type, so as to process the received voice instruction through the target voice recognition model.

[0108] Regarding the above-mentioned device, the specific functions of each module therein have been described in detail in the embodiment of the voice wake-up method provided in the embodiment of this specification, and will not be elaborated here.

[0109] Based on the same inventive concept, an embodiment of the present invention provides an electronic device, referring to Figure 4 As shown, it includes: a memory 404, a processor 402, and a computer program stored in the memory 404 and executable on the processor 402. When the processor 402 executes the program, any implementation of the voice wake-up method embodiment is implemented.

[0110] Among them, Figure 4In the embodiment of the present invention, a bus architecture (represented by bus 400) is shown, which may include any number of interconnected buses and bridges, and bus 400 links various circuits including one or more processors represented by processor 402 and memory represented by memory 404. Bus 400 may also link various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. Bus interface 405 provides an interface between bus 400 and receiver 401 and transmitter 403. Receiver 401 and transmitter 403 may be the same element, namely a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 402 is responsible for managing bus 400 and general processing, while memory 404 may be used to store data used by processor 402 when performing operations.

[0111] The functions described herein may be implemented in hardware, software executed by a processor, firmware, or any combination thereof. If implemented in software executed by a processor, the functions may be stored as one or more instructions or codes on a computer-readable medium or transmitted via a computer-readable medium. Other examples and implementations are within the scope and spirit of the present invention and the appended claims. For example, due to the nature of software, the functions described above may be implemented using software executed by a processor, hardware, firmware, hard wiring, or a combination of any of these. In addition, each functional unit may be integrated into a processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit.

[0112] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0113] The units described as separate components may or may not be physically separated, and the components of the control device may or may not be physical units, that is, they may be located in one place or distributed in multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0114] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0115] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the claims of the present invention.

Claims

1. A voice wake-up method, characterized in that: include: If target wake-up audio information is detected, performing accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information; Determining a target accent awakening model corresponding to the target accent type based on a correspondence between the accent type and the accent awakening model; The target wake-up audio information is processed based on the target accent wake-up model to perform a wake-up operation.

2. The method according to claim 1, characterized in that Before the target wake-up audio information is detected and the accent type is recognized for the target wake-up audio information, the method further includes: receiving first audio information; The first audio information is matched with the reference wake-up audio information. If the match is successful, it is determined that the target wake-up audio information is detected, and the reference wake-up audio information is the audio information corresponding to the wake-up word.

3. The method according to claim 1, characterized in that The performing accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information includes: Based on a preset accent classification model, the target wake-up audio is identified to obtain the target accent type corresponding to the target wake-up audio, wherein the target accent type is a standard accent type, a dialect accent type, or a pronunciation deviation accent type.

4. The method according to claim 1, characterized in that Each accent awakening model in the corresponding relationship is obtained by the following steps: For each accent wake-up model, an initial wake-up model is constructed and a training sample set is obtained, wherein each sample in the training sample set is audio information under a corresponding accent type; Based on the training sample set under each accent type, the corresponding initial awakening model is trained to obtain a trained accent awakening model.

5. The method according to claim 4, characterized in that The step of obtaining a training sample set includes: Acquiring second audio information; Matching the second audio information with each piece of audio information in a preset accent library, wherein the preset accent library includes a plurality of preset audio information with accent tags, each accent tag corresponding to an accent type; If the matching result shows that the similarity between the second audio information and the N preset audio information in the preset accent library is greater than or equal to a threshold, the accent tag of the preset audio information corresponding to the maximum similarity is used as the accent tag of the second audio information, where N is a positive integer; Based on the accent label of the second audio information, the second audio information is divided into a training sample set under a corresponding accent type.

6. The method according to claim 5, characterized in that After matching the second audio information with each piece of audio information in the preset accent library, the method further includes: If the matching result indicates that the similarity between the second audio information and each piece of preset audio information in the preset accent library is less than the threshold, outputting update confirmation information for the preset accent library, wherein the update confirmation information is used to determine whether the accent tag of the second audio information is a newly added accent tag; receiving feedback information in response to the update confirmation information; If the feedback information confirms that the accent tag of the second audio information is a newly added accent tag, the accent tag corresponding to the second audio information is obtained, and the second audio information and the corresponding accent tag are added to the preset accent library.

7. The method according to claim 1, characterized in that After performing accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information, the method further includes: A target speech recognition model corresponding to the target accent type is determined to process the received voice instruction through the target speech recognition model.

8. A voice wake-up device, characterized in that: include: An accent recognition module, configured to, if target wake-up audio information is detected, perform accent type recognition on the target wake-up audio information to obtain a target accent type corresponding to the target wake-up audio information; A model determination module, configured to determine a target accent awakening model corresponding to the target accent type based on a correspondence between the accent type and the accent awakening model; A processing module is used to process the target wake-up audio information based on the target accent wake-up model to perform a wake-up operation.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images