Voice keyword detection method and device, equipment and storage medium

Abstract

The embodiment of the invention provides a voice keyword detection method and device, electronic equipment and a storage medium, belongs to the technical field of voice processing, and is suitable for the field of financial science and technology. The method comprises the following steps: acquiring voice sample data; obtaining an initial feature extraction sub-model and an initial confrontation classification sub-model; on the basis of the initial feature extraction sub-model, keyword detection is carried out on the voice sample data to obtain predicted keywords; calculating features and extracting loss data based on the predicted keywords and the key frame marks; based on the initial adversarial classification sub-model and the sample voice features, performing type recognition on the voice sample data to obtain a voice prediction category; calculating adversarial classification loss data based on the voice prediction category and the voice tag; and performing gradient inversion on the initial feature extraction sub-model and the initial adversarial classification sub-model based on the feature extraction loss data and the adversarial classification loss data. According to the embodiment of the invention, the cost consumption of voice keyword detection can be reduced.

Description

Technical Field

[0001] The present application relates to the field of speech processing technology, is applicable to the field of financial technology, and in particular to a speech keyword detection method and device, an electronic device, and a storage medium. Background Art

[0002] Voice keyword detection refers to the detection and extraction of keywords in voice data. For example, in the financial field, voice keyword detection can be performed on the voice data of communication between financial customer service and users to help financial customer service quickly review the communication content between them and users.

[0003] At present, the common method of voice keyword detection is usually to use neural networks to detect keywords in voice data. However, training neural networks requires a large amount of voice data and is costly. Therefore, how to reduce the cost of voice keyword detection has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to provide a voice keyword detection method and device, an electronic device and a storage medium, aiming to reduce the cost of voice keyword detection.

[0005] To achieve the above object, a first aspect of an embodiment of the present application provides a method for detecting speech keywords, the method comprising:

[0006] Acquire speech sample data including speech tags and key frame tags, wherein the speech sample data includes real speech data and synthesized speech data, the speech tags are used to indicate that the speech sample data is real speech data or synthesized speech data, and the key frame tags are used to identify speech frames in the speech sample data;

[0007] Acquire an initial keyword detection model, wherein the initial keyword detection model includes an initial feature extraction sub-model and an initial adversarial classification sub-model;

[0008] Based on the initial feature extraction sub-model, feature extraction is performed on the speech sample data to obtain sample speech features;

[0009] Based on the initial feature extraction sub-model and the sample speech features, performing keyword detection on the speech sample data to obtain predicted keywords;

[0010] Calculating feature extraction loss data based on the predicted keywords and the key frame markers;

[0011] Based on the initial adversarial classification sub-model and the sample speech features, the type of the speech sample data is identified to obtain a speech prediction category, wherein the speech prediction category is used to indicate that the speech sample data is the real speech data or the synthesized speech data;

[0012] Calculating adversarial classification loss data based on the speech prediction category and the speech label;

[0013] Based on the feature extraction loss data and the adversarial classification loss data, performing gradient reversal on the initial keyword detection model to obtain a target keyword detection model;

[0014] Based on the target keyword detection model, keyword detection is performed on the target voice data to obtain the target keyword.

[0015] In some embodiments, performing gradient reversal on the initial keyword detection model based on the feature extraction loss data and the adversarial classification loss data to obtain a target keyword detection model includes:

[0016] Calculating feature extraction gradient data based on the feature extraction loss data and preset model parameters, wherein the model parameters refer to parameters in the initial keyword detection model;

[0017] Calculating adversarial classification gradient data based on the adversarial classification loss data and the model parameters;

[0018] Performing data inversion on the adversarial classification gradient data to obtain inverted gradient data;

[0019] Based on the feature extraction gradient data, the adversarial classification gradient data and the inversion gradient data, the initial keyword detection model is back-propagated to obtain the target keyword detection model.

[0020] In some embodiments, the back-propagating the initial keyword detection model based on the feature extraction gradient data, the adversarial classification gradient data, and the reverse gradient data to obtain the target keyword detection model includes:

[0021] Based on the adversarial classification gradient data and the feature extraction gradient data, adjusting parameters of the initial adversarial classification sub-model to obtain a target adversarial classification sub-model;

[0022] Based on the feature extraction gradient data and the inverted gradient data, adjusting parameters of the initial feature extraction sub-model to obtain a target feature extraction sub-model;

[0023] The target feature extraction sub-model and the target adversarial classification sub-model are concatenated to obtain the target keyword detection model.

[0024] In some embodiments, the acquiring of speech sample data including speech tags and key frame tags includes:

[0025] Get recorded voice data;

[0026] Performing data identification on the recorded voice data to obtain the real voice data, wherein the real voice data includes a real label and a real key frame label, the real label is used to indicate that the artificially recorded data is the real voice data, and the real key frame label is used to identify the voice frame in the real voice data;

[0027] Performing voice simulation on the real voice data to obtain simulated voice data;

[0028] Performing data identification on the simulated speech data to obtain the synthesized speech data, wherein the synthesized speech data includes a synthesis tag and a synthesis key frame tag, the synthesis tag is used to indicate that the simulated speech data is the synthesized speech data, and the synthesis key frame tag is used to identify a speech frame in the synthesized speech data;

[0029] The real voice data and the synthesized voice data are merged to obtain the voice sample data.

[0030] In some embodiments, performing voice simulation on the real voice data to obtain simulated voice data includes:

[0031] Performing voice transcription on the real voice data to obtain voice text data;

[0032] Performing data enhancement on the speech text data to obtain enhanced text data;

[0033] Speech synthesis is performed based on the enhanced text data to obtain the simulated speech data.

[0034] In some embodiments, performing keyword detection on target speech data based on the target keyword detection model to obtain target keywords includes:

[0035] Based on the target keyword detection model, feature extraction is performed on the target speech data to obtain target speech features;

[0036] Based on the target keyword detection model and the target speech features, performing context understanding on the target speech data to obtain context information;

[0037] Based on the context information, feature classification is performed on the target speech feature to obtain a target feature category;

[0038] Based on the target feature category, linearly mapping the target speech feature to obtain a keyword probability, wherein the keyword probability is used to indicate the probability that the target speech feature is a keyword;

[0039] Based on the keyword probability, the target speech feature is screened to obtain the target keyword.

[0040] In some embodiments, the performing type recognition on the speech sample data based on the initial adversarial classification sub-model and the sample speech features to obtain a speech prediction category includes:

[0041] Based on the initial adversarial classification sub-model, feature recognition is performed on the sample speech features to obtain sample feature categories;

[0042] Based on the sample feature category, data tracing is performed on the sample speech feature to obtain the speech prediction category.

[0043] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a speech keyword detection device, the device comprising:

[0044] A sample data acquisition module, used to acquire speech sample data including speech tags and key frame tags, wherein the speech sample data includes real speech data and synthesized speech data, the speech tags are used to indicate that the speech sample data is real speech data or synthesized speech data, and the key frame tags are used to identify speech frames in the speech sample data;

[0045] An initial model acquisition module, used to acquire an initial keyword detection model, wherein the initial keyword detection model includes an initial feature extraction sub-model and an initial adversarial classification sub-model;

[0046] A speech feature extraction module, used for performing feature extraction on the speech sample data based on the initial feature extraction sub-model to obtain sample speech features;

[0047] A keyword prediction module, used to perform keyword detection on the speech sample data based on the initial feature extraction sub-model and the sample speech features to obtain predicted keywords;

[0048] A first loss calculation module, configured to calculate feature extraction loss data based on the predicted keywords and the key frame markers;

[0049] A speech data classification module, used for performing type recognition on the speech sample data based on the initial adversarial classification sub-model and the sample speech features to obtain a speech prediction category, wherein the speech prediction category is used to indicate that the speech sample data is the real speech data or the synthesized speech data;

[0050] A second loss calculation module, used to calculate adversarial classification loss data based on the speech prediction category and the speech label;

[0051] A model parameter adjustment module is used to perform gradient inversion on the initial keyword detection model based on the feature extraction loss data and the adversarial classification loss data to obtain a target keyword detection model.

[0052] The keyword detection module is used to perform keyword detection on the target speech data based on the target keyword detection model to obtain the target keyword.

[0053] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0054] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0055] The speech keyword detection method and device, electronic device and storage medium proposed in the present application perform feature extraction on speech sample data through an initial feature extraction submodel to obtain sample speech features, then perform keyword detection on the speech sample data based on the initial feature extraction submodel and the sample speech features to obtain predicted keywords, and calculate feature extraction loss data based on the predicted keywords and key frame tags, so as to realize keyword detection on real speech data and synthetic speech data. Furthermore, the type of speech sample data is identified through an initial adversarial classification submodel and sample speech features to obtain speech prediction categories, and then adversarial classification loss data is calculated based on the speech prediction categories and speech tags, and the initial keyword detection model is gradient reversed based on the feature extraction loss data and the adversarial classification loss data, so that the speech features extracted by the initial feature extraction submodel are more inclined to the common features of synthetic speech data and real speech data, thereby reducing the demand for real speech data for training the keyword detection model, thereby reducing the cost consumption of speech keyword detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1is a flow chart of a method for detecting speech keywords provided in an embodiment of the present application;

[0057] Figure 2 yes Figure 1 Flow chart of step S101 in FIG.

[0058] Figure 3 yes Figure 2 Flow chart of step S203 in FIG.

[0059] Figure 4 yes Figure 1 Flow chart of step S106 in FIG.

[0060] Figure 5 yes Figure 1 Flow chart of step S108 in FIG.

[0061] Figure 6 yes Figure 5 Flow chart of step S504 in FIG.

[0062] Figure 7 yes Figure 1 Flow chart of step S109 in FIG.

[0063] Figure 8 It is a structural diagram of a speech keyword detection device provided in an embodiment of the present application;

[0064] Fig. 9 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0066] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0067] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0068] First, some nouns involved in this application are analyzed:

[0069] Virtuso model: The Virtuso model aims to achieve efficient processing and generation of voice data by combining language models and audio feature extraction technology, supporting a variety of application scenarios such as speech recognition, music creation, and sound synthesis, so as to improve the naturalness and fluency of voice interaction.

[0070] Audio Language Model: An audio language model is a model for audio generation and understanding that aims to generate coherent audio clips from existing audio data.

[0071] Voice keyword detection refers to the detection and extraction of keywords in voice data. For example, in the financial field, voice keyword detection can be performed on the voice data of communication between financial customer service and users to help financial customer service quickly review the communication content between them and users.

[0072] At present, the common method of voice keyword detection is usually to use neural networks to detect keywords in voice data. However, training neural networks requires a large amount of voice data and is costly. Therefore, how to reduce the cost of voice keyword detection has become a technical problem that needs to be solved urgently.

[0073] Based on this, the embodiments of the present application provide a voice keyword detection method and device, an electronic device and a storage medium, aiming to reduce the cost of voice keyword detection.

[0074] The voice keyword detection method and device, electronic device and storage medium provided in the embodiments of the present application are specifically described through the following embodiments. First, the voice keyword detection method in the embodiments of the present application is described.

[0075] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0076] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0077] The voice keyword detection method provided in the embodiment of the present application relates to the field of voice processing technology and is applicable to the field of financial technology. The voice keyword detection method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the voice keyword detection method, etc., but is not limited to the above forms.

[0078] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0079] Figure 1 is an optional flow chart of the voice keyword detection method provided in the embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S109.

[0080] Step S101, obtaining speech sample data including speech tags and key frame tags, wherein the speech sample data includes real speech data and synthesized speech data, the speech tags are used to indicate whether the speech sample data is real speech data or synthesized speech data, and the key frame tags are used to identify speech frames in the speech sample data;

[0081] Step S102, obtaining an initial keyword detection model, wherein the initial keyword detection model includes an initial feature extraction sub-model and an initial adversarial classification sub-model;

[0082] Step S103, based on the initial feature extraction sub-model, feature extraction is performed on the speech sample data to obtain sample speech features;

[0083] Step S104, based on the initial feature extraction sub-model and the sample speech features, keyword detection is performed on the speech sample data to obtain predicted keywords;

[0084] Step S105, calculating feature extraction loss data based on the predicted keywords and key frame tags;

[0085] Step S106, based on the initial adversarial classification sub-model and the sample speech features, the type of the speech sample data is identified to obtain a speech prediction category, wherein the speech prediction category is used to indicate whether the speech sample data is real speech data or synthesized speech data;

[0086] Step S107, calculating adversarial classification loss data based on the speech prediction category and the speech label;

[0087] Step S108, based on the feature extraction loss data and the adversarial classification loss data, gradient inversion is performed on the initial keyword detection model to obtain a target keyword detection model;

[0088] Step S109: Based on the target keyword detection model, keyword detection is performed on the target voice data to obtain the target keyword.

[0089] In steps S101 to S109 shown in the embodiment of the present application, feature extraction is performed on the acquired speech sample data through the acquired initial feature extraction sub-model to obtain sample speech features, and then keyword detection is performed on the speech sample data based on the initial feature extraction sub-model and the sample speech features to obtain predicted keywords, and feature extraction loss data is calculated based on the predicted keywords and key frame labels. Secondly, type recognition is performed on the speech sample data through the acquired initial adversarial classification sub-model and the above-mentioned sample speech features to obtain a speech prediction category indicating that the speech sample data is real speech data or synthetic speech data, and adversarial classification loss data is calculated based on the speech prediction category and the speech label. Finally, based on the feature extraction loss data and the adversarial classification loss data, gradient inversion is performed on the initial keyword detection model to obtain a target keyword detection model, and keyword detection is performed on the target speech data based on the target keyword detection model to obtain the target keyword. Therefore, the present application uses an initial feature extraction sub-model to perform feature extraction on speech sample data to obtain sample speech features, and then performs keyword detection on the speech sample data based on the initial feature extraction sub-model and the sample speech features to obtain predicted keywords, and calculates feature extraction loss data based on the predicted keywords and key frame labels, thereby realizing keyword detection on real speech data and synthetic speech data. Furthermore, through the initial adversarial classification sub-model and sample speech features, type recognition is performed on the speech sample data to obtain speech prediction categories, and then adversarial classification loss data is calculated based on the speech prediction categories and speech labels, and the initial keyword detection model is gradient reversed based on the feature extraction loss data and the adversarial classification loss data, so that the speech features extracted by the initial feature extraction sub-model are more inclined to the common features of synthetic speech data and real speech data, thereby reducing the demand for real speech data for training the keyword detection model, thereby reducing the cost consumption of speech keyword detection.

[0090] In step S101 of some embodiments, the real voice data may be voice data obtained by artificial voice recording, and the synthesized voice data may be voice data generated based on voice synthesis technology.

[0091] The present application can obtain real voice data by manually recording voice, and then perform data identification on the real voice data to obtain real labels and real key frame labels of the real voice data, and then perform voice simulation based on the real voice data to generate simulated voice data, and further perform synthetic voice identification and key frame identification on the simulated voice data to obtain synthetic voice data containing synthetic labels and synthetic key frame labels, and finally, merge the above real voice data and synthetic voice data to obtain voice sample data.

[0092] In detail, in some embodiments, step S101 may include but is not limited to steps S201 to S205:

[0093] Step S201, obtaining recorded voice data;

[0094] Step S202, data identification is performed on the recorded voice data to obtain real voice data, wherein the real voice data includes a real label and a real key frame label, the real label is used to indicate that the artificially recorded data is real voice data, and the real key frame label is used to identify the voice frame in the real voice data;

[0095] Step S203, performing voice simulation on the real voice data to obtain simulated voice data;

[0096] Step S204, performing data identification on the simulated speech data to obtain synthesized speech data, wherein the synthesized speech data includes a synthesis tag and a synthesis key frame tag, the synthesis tag is used to indicate that the simulated speech data is synthesized speech data, and the synthesis key frame tag is used to identify speech frames in the synthesized speech data;

[0097] Step S205: merge the real speech data and the synthesized speech data to obtain speech sample data.

[0098] In step S201 and step S202 of some embodiments, recorded voice data can be obtained by acquiring artificially recorded voice data, and then the recorded voice data is labeled to obtain real labels for indicating that the recorded voice data is real voice data, as well as real key frame labels for identifying key voice frames in the real voice data.

[0099] In step S203 of some embodiments, the simulated voice data can be obtained by converting the real voice data into voice text data, modifying the voice text data to obtain a plurality of enhanced text data, and then converting the enhanced text data into voice data.

[0100] In detail, in some embodiments, step S203 may include but is not limited to steps S301 to S303:

[0101] Step S301, performing voice transcription on the real voice data to obtain voice text data;

[0102] Step S302, performing data enhancement on the speech text data to obtain enhanced text data;

[0103] Step S303, performing speech synthesis based on the enhanced text data to obtain simulated speech data.

[0104] In step S301 of some embodiments, the real speech data is subjected to audio data preprocessing such as noise elimination and audio segmentation to obtain standard speech data, and feature extraction is performed on the standard speech data to convert the audio data into a feature vector. The feature vector is then input into a trained long short-term memory network, a convolutional neural network, or a deep learning model based on a converter architecture, which can achieve speech transcription of the real speech data to obtain speech text data.

[0105] In step S302 of some embodiments, data enhancement of the voice text data can be achieved by replacing the key words in the voice text data, thereby obtaining enhanced text data. For example, in a scenario where a salesperson promotes insurance business, phrases such as "return on investment" and "diversified investment" that appear in the voice text data can be replaced with synonyms or synonyms, thereby achieving data enhancement of the voice text data, thereby obtaining enhanced text data.

[0106] It should be noted that in the embodiments of the present application, the real speech data can also be placed in a model with speech transcription and data enhancement for processing. For example, the Virtuso model can be used to perform speech transcription and data enhancement on the real speech data.

[0107] In step S303 of some embodiments, an audio generation model based on a language model may be used to generate simulated speech data according to the enhanced text data. Specifically, the audio generation model may be an AudioLM (Audio Language Model).

[0108] In steps S301 to S303 shown in this embodiment, by performing speech transcription on real speech data to obtain speech text data, performing data enhancement on the speech text data to obtain enhanced text data, and then performing speech synthesis based on the enhanced text data to obtain simulated speech data, the demand for real speech data can be reduced, thereby reducing the cost of model training.

[0109] In step S204 of some embodiments, by labeling the simulated speech data, it is possible to obtain a label indicating that the simulated speech data is synthesized speech data, and by labeling the speech frames of the simulated speech data, it is possible to obtain a synthesized key frame label for identifying key speech frames in the synthesized speech data. Through the above-mentioned synthesized label and synthesized key frame label, the simulated speech data can be converted into synthesized speech data containing the synthesized label and the synthesized key frame label.

[0110] In step S205 of some embodiments, the real speech data and the synthesized speech data are combined into one data set to obtain speech sample data.

[0111] In steps S201 to S205 shown in the present embodiment, the acquired recorded voice data is data-labeled to obtain real voice data including real labels and real key frame labels, and then the real voice data is voice simulated to obtain simulated voice data. Furthermore, the simulated voice data is data-labeled to obtain synthetic voice data including synthetic labels and synthetic key frame labels, and then the real voice data and the synthetic voice data are data-merged to obtain voice sample data, which can reduce the demand for real voice data in the training model, thereby reducing the cost of model training.

[0112] In step S102 of some embodiments, the initial keyword detection model is used to perform keyword detection on the input voice data. Specifically, the initial keyword detection model can be divided into an initial feature extraction sub-model and an initial adversarial classification sub-model, wherein the initial feature extraction sub-model is used to extract voice features in the voice data, and then perform keyword extraction based on the voice features, and the initial adversarial classification sub-model is used to determine whether the input voice data is real voice data or synthesized voice data based on the voice features extracted by the initial feature extraction sub-model.

[0113] The present application can connect the input of the initial adversarial classification sub-model with the feature extraction output interface of the initial feature extraction sub-model, thereby realizing the use of the speech features extracted by the initial feature extraction sub-model as the input data of the initial adversarial classification sub-model.

[0114] In step S103 of some embodiments, the present application can convert the speech sample data into a Mel spectrum, and then determine the sample speech features of the speech sample data based on the amplitude, frequency and other features of the Mel spectrum, thereby realizing feature extraction of the speech sample data using the initial feature extraction sub-model.

[0115] In step S104 of some embodiments, the initial feature extraction submodel can be used to perform contextual understanding on the sample speech features extracted above to obtain contextual information of the sample speech features, and then the sample speech features are classified into keywords in combination with the contextual information to obtain speech feature categories, and then a pre-set activation function is used to linearly map the speech feature categories to obtain the keyword sample probability indicating that the sample speech features are keywords, and finally, according to the keyword sample probability, some features are screened out from the sample speech features as predicted keywords.

[0116] In step S105 of some embodiments, speech transcription is performed on the area marked by the key frame in the speech sample data to obtain the actual speech keywords, and then the predicted keywords are compared with the actual speech keywords to obtain feature extraction loss data representing the gap between the predicted keywords and the actual speech keywords.

[0117] In step S106 of some embodiments, the category of the sample speech feature is identified using the initial adversarial classification sub-model to obtain the sample feature category, and then based on the identified sample feature category, the sample speech feature is traced back to obtain the type classification of each speech sample data, that is, the speech prediction category.

[0118] For details, see Figure 4 In some embodiments, step S106 may include but is not limited to steps S401 to S402:

[0119] Step S401, based on the initial adversarial classification sub-model, feature recognition is performed on the sample speech features to obtain the sample feature category;

[0120] Step S402: based on the sample feature category, data tracing is performed on the sample speech feature to obtain the speech prediction category.

[0121] In step S401 of some embodiments, since there are similarities and differences between the synthesized speech data and the real speech data, in the initial stage of model training, the initial feature extraction submodel often extracts the differences between the two. Therefore, the initial adversarial classification submodel can perform feature classification on the input sample speech features according to the differences between the synthesized speech data and the real speech data learned in advance, thereby obtaining the sample feature categories.

[0122] In step S402 of some embodiments, after obtaining the sample feature category, the speech sample data can be classified by tracing back the speech sample data represented by the sample speech feature, thereby obtaining the speech prediction category.

[0123] In steps S401 to S402 shown in this embodiment, the initial adversarial classification sub-model can be used to perform feature recognition on the sample speech features to obtain the sample feature categories, and then the sample speech features can be traced according to the sample feature categories, so as to realize the type recognition of the speech sample data, thereby helping the initial keyword extraction model to distinguish between the synthesized speech data and the real speech data.

[0124] In step S107 of some embodiments, the actual speech category of the speech sample data can be clarified through the speech labels in the speech sample data, and the adversarial classification loss data of the initial adversarial classification sub-model can be obtained by calculating the difference between the predicted speech category and the actual speech category.

[0125] In step S108 of some embodiments, feature extraction is performed by calculating feature extraction loss data and adversarial classification loss data during back propagation of the initial keyword detection model, and then the gradient is reversed and passed to the initial feature extraction sub-model, so that the initial feature extraction sub-model and the initial adversarial classification sub-model form an adversarial relationship, thereby increasing the probability that the initial feature extraction sub-model extracts common features between the synthetic speech data and the real speech data, making it difficult for the initial adversarial classification sub-model to distinguish the speech category of the speech sample data, thereby realizing the use of synthetic speech data instead of real speech data for training the initial keyword detection model.

[0126] It is important to know that this can be accomplished by adding a gradient reversal layer to the initial keyword detection model.

[0127] For details, see Figure 5 In some embodiments, step S108 may include but is not limited to steps S501 to S504:

[0128] Step S501, calculating feature extraction gradient data based on feature extraction loss data and preset model parameters, wherein the model parameters refer to parameters in the initial keyword detection model;

[0129] Step S502, calculating adversarial classification gradient data based on adversarial classification loss data and model parameters;

[0130] Step S503, performing data inversion on the adversarial classification gradient data to obtain inverted gradient data;

[0131] Step S504, based on the feature extraction gradient data, the adversarial classification gradient data and the inversion gradient data, the initial keyword detection model is back-propagated to obtain a target keyword detection model.

[0132] In steps S501 and S502 of some embodiments, based on the feature extraction loss data and the adversarial classification loss data, the change rates of the feature extraction loss function and the adversarial classification loss function under the current parameter values ​​can be obtained for each parameter in the initial keyword detection model, that is, the feature extraction gradient data and the adversarial classification gradient data, wherein the feature extraction loss function is a function that calculates the feature extraction loss data based on predicted keywords and key frame markers, and the adversarial classification loss function is a function that calculates the adversarial classification loss data based on the speech prediction category and speech label.

[0133] In step S503 of some embodiments, by inverting the above-mentioned adversarial classification gradient data, reverse adversarial classification gradient data, i.e., reversed gradient data, can be obtained so that the initial feature extraction submodule can make parameter adjustments to counter the initial adversarial classification submodel according to the reversed gradient data.

[0134] In step S504 of some embodiments, the feature extraction gradient data, adversarial classification gradient data and reverse gradient data are back-propagated to the initial feature extraction sub-model and the initial adversarial classification sub-model, so that the parameters of the initial feature extraction sub-model and the initial adversarial classification sub-model are adjusted, thereby obtaining a target keyword detection model.

[0135] For details, see Figure 6 In some embodiments, step S504 may include but is not limited to steps S601 to S603:

[0136] Step S601, adjusting the parameters of the initial adversarial classification sub-model based on the adversarial classification gradient data and the feature extraction gradient data to obtain the target adversarial classification sub-model;

[0137] Step S602, adjusting parameters of the initial feature extraction sub-model based on the feature extraction gradient data and the reverse gradient data to obtain a target feature extraction sub-model;

[0138] Step S603, concatenating the target feature extraction sub-model and the target adversarial classification sub-model to obtain a target keyword detection model.

[0139] In step S601 of some embodiments, the adversarial classification gradient data and the feature extraction gradient data are combined to obtain forward combined gradient data, and then the parameters of the initial adversarial classification sub-model are adjusted based on the forward combined gradient data. This enables the initial adversarial classification sub-model to more accurately identify the category of the speech sample data, thereby obtaining the target adversarial classification sub-model.

[0140] In step S602 of some embodiments, the inverted gradient data and the feature extraction gradient data are combined to obtain reverse combined gradient data, and then the parameters of the initial feature extraction sub-model are adjusted based on the reverse combined gradient data, so that the initial feature extraction model produces a trend that prevents the initial adversarial classification sub-model from enhancing the ability to distinguish the category of speech sample data, thereby helping the initial feature extraction model to extract common features between the synthesized speech data and the real speech data to form a target feature extraction sub-model.

[0141] In step S603 of some embodiments, the target feature extraction sub-model and the target adversarial classification sub-model are concatenated to obtain a trained target keyword detection model.

[0142] In steps S601 to S603 shown in the present embodiment, parameters of the initial adversarial classification sub-model are adjusted according to the adversarial classification gradient data and the feature extraction gradient data to obtain a target adversarial classification sub-model, and then parameters of the initial feature extraction sub-model are adjusted according to the feature extraction gradient data and the reverse gradient data to obtain a target feature extraction sub-model. Finally, the target feature extraction sub-model and the target adversarial classification sub-model are concatenated to obtain a target keyword detection model, thereby reducing the demand for real speech data during the training process of the target keyword detection model and achieving a reduction in cost consumption during the training process of the target keyword detection model.

[0143] In steps S501 to S504 shown in the present embodiment, feature extraction gradient data is calculated based on feature extraction loss data and model parameters. At the same time, adversarial classification gradient data is calculated based on adversarial classification loss data and model parameters. Furthermore, the adversarial classification gradient data is inverted to obtain inverted gradient data. Then, based on the feature extraction gradient data, adversarial classification gradient data and inverted gradient data, the initial keyword detection model is back-propagated to obtain a target keyword detection model, which enables the initial feature extraction sub-model to form an adversarial relationship with the initial adversarial classification sub-model, thereby increasing the probability that the initial feature extraction sub-model extracts common features between the synthesized speech data and the real speech data.

[0144] In step S109 of some embodiments, the target voice data may be voice data that has not been tagged and key-framed. In an insurance sales conversation, the target voice data may be real-time recording data of a salesperson.

[0145] The present application can utilize the above-mentioned trained target keyword detection model to perform feature extraction, feature context understanding, feature classification and other processing on the target speech data, thereby obtaining the target keywords in the target speech data.

[0146] For details, see Figure 7 In some embodiments, before step S106, the voice keyword detection method may also include but is not limited to steps S701 to S705:

[0147] Step S701, based on the target keyword detection model, feature extraction is performed on the target speech data to obtain target speech features;

[0148] Step S702, based on the target keyword detection model and the target speech features, the target speech data is contextually understood to obtain context information;

[0149] Step S703, based on the context information, feature classification is performed on the target speech feature to obtain a target feature category;

[0150] Step S704, based on the target feature category, linear mapping is performed on the target speech feature to obtain a keyword probability, wherein the keyword probability is used to indicate the probability that the target speech feature is a keyword;

[0151] Step S705: based on the keyword probability, the target speech features are screened to obtain the target keywords.

[0152] In step S701 of some embodiments, based on the target keyword detection model, feature extraction is performed on the target speech data to obtain the target speech feature. This method is similar to the above-mentioned method of "based on the initial feature extraction sub-model, feature extraction is performed on the speech sample data to obtain the sample speech feature", so it is not repeated here.

[0153] In step S702 of some embodiments, by performing vector product calculations on the vocabulary vector in the context of the target speech feature in the room exchange target speech data and the target speech feature, the relationship between the target speech feature and the context of the target speech feature can be clarified, thereby obtaining the context information of the target speech feature.

[0154] In step S703 of some embodiments, semantic understanding of the target speech feature is performed based on the context information of the target speech feature, and the feature category of the target speech feature can be clarified, thereby achieving feature classification of the target speech feature and obtaining the target feature category.

[0155] In step S704 and step S705 of some embodiments, according to the target feature category, the target speech feature is linearly mapped through the activation function pre-set in the target keyword detection model to obtain the keyword probability that the target speech feature is the target keyword, and then based on the keyword probability, the target speech features that meet the preset conditions are screened out from the target speech features for output, thereby obtaining the target keyword.

[0156] In steps S701 to S705 shown in the present embodiment, feature extraction is performed on the target speech data according to the target keyword detection model to obtain target speech features, and then context understanding is performed on the target speech data according to the target keyword detection model and the target speech features to obtain context information. Furthermore, feature classification is performed on the target speech features according to the context information to obtain target feature categories. Next, linear mapping is performed on the target speech features according to the target feature categories to obtain keyword probabilities, wherein the keyword probability is used to indicate the probability that the target speech feature is a keyword. Finally, the target speech features are screened according to the keyword probability, so that keyword extraction of the target speech data can be realized quickly and conveniently, thereby improving the efficiency of speech keyword detection.

[0157] The present application uses an initial feature extraction sub-model to perform feature extraction on speech sample data to obtain sample speech features, and then performs keyword detection on the speech sample data based on the initial feature extraction sub-model and the sample speech features to obtain predicted keywords, and calculates feature extraction loss data based on the predicted keywords and key frame tags, thereby realizing keyword detection on real speech data and synthetic speech data. Furthermore, the application uses an initial adversarial classification sub-model and sample speech features to perform type recognition on speech sample data to obtain speech prediction categories, and then calculates adversarial classification loss data based on the speech prediction categories and speech tags, and performs gradient inversion on the initial keyword detection model based on the feature extraction loss data and the adversarial classification loss data, thereby making the speech features extracted by the initial feature extraction sub-model more similar to the common features of synthetic speech data and real speech data, thereby reducing the demand for real speech data for training the keyword detection model, thereby reducing the cost consumption of speech keyword detection.

[0158] See also Figure 8 The present application also provides a speech keyword detection device, which can implement the above-mentioned speech keyword detection method, and the device includes:

[0159] The sample data acquisition module 801 is used to acquire speech sample data including speech tags and key frame tags, wherein the speech sample data includes real speech data and synthesized speech data, the speech tags are used to indicate whether the speech sample data is real speech data or synthesized speech data, and the key frame tags are used to identify speech frames in the speech sample data;

[0160] An initial model acquisition module 802 is used to acquire an initial keyword detection model, wherein the initial keyword detection model includes an initial feature extraction sub-model and an initial adversarial classification sub-model;

[0161] The speech feature extraction module 803 is used to extract features from the speech sample data based on the initial feature extraction sub-model to obtain sample speech features;

[0162] The keyword prediction module 804 is used to perform keyword detection on the speech sample data based on the initial feature extraction sub-model and the sample speech features to obtain predicted keywords;

[0163] A first loss calculation module 805, for calculating feature extraction loss data based on predicted keywords and key frame tags;

[0164] A speech data classification module 806 is used to identify the type of speech sample data based on the initial adversarial classification sub-model and the sample speech features to obtain a speech prediction category, wherein the speech prediction category is used to indicate whether the speech sample data is real speech data or synthesized speech data;

[0165] A second loss calculation module 807, used to calculate adversarial classification loss data based on the speech prediction category and the speech label;

[0166] Model parameter adjustment module 808, used to perform gradient inversion on the initial keyword detection model based on feature extraction loss data and adversarial classification loss data to obtain a target keyword detection model

[0167] The keyword detection module 809 is used to perform keyword detection on the target speech data based on the target keyword detection model to obtain the target keyword.

[0168] The specific implementation of the voice keyword detection device is basically the same as the specific embodiment of the voice keyword detection method described above, and will not be repeated here.

[0169] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned voice keyword detection method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0170] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0171] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0172] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the voice keyword detection method of the embodiment of this application;

[0173] Input / output interface 903, used to implement information input and output;

[0174] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0175] A bus 905 that transmits information between the various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0176] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0177] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned voice keyword detection method is implemented.

[0178] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0179] The speech keyword detection method, speech keyword detection device, electronic device and storage medium provided in the embodiments of the present application perform feature extraction on the acquired speech sample data through the acquired initial feature extraction submodel to obtain sample speech features, then perform keyword detection on the speech sample data according to the initial feature extraction submodel and the sample speech features to obtain predicted keywords, and calculate feature extraction loss data according to the predicted keywords and key frame tags. Secondly, perform type recognition on the speech sample data through the acquired initial adversarial classification submodel and the above-mentioned sample speech features to obtain a speech prediction category indicating that the speech sample data is real speech data or synthetic speech data, and calculate adversarial classification loss data according to the speech prediction category and the speech label. Finally, perform gradient inversion on the initial keyword detection model according to the feature extraction loss data and the adversarial classification loss data to obtain a target keyword detection model, and perform keyword detection on the target speech data according to the target keyword detection model to obtain the target keyword.

[0180] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0181] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0182] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0183] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0184] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application 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 interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0185] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0186] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0187] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0188] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0189] 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 application, 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 multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0190] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for detecting speech keywords, characterized in that: The method comprises: Acquire speech sample data including speech tags and key frame tags, wherein the speech sample data includes real speech data and synthesized speech data, the speech tags are used to indicate that the speech sample data is real speech data or synthesized speech data, and the key frame tags are used to identify speech frames in the speech sample data; Acquire an initial keyword detection model, wherein the initial keyword detection model includes an initial feature extraction sub-model and an initial adversarial classification sub-model; Based on the initial feature extraction sub-model, feature extraction is performed on the speech sample data to obtain sample speech features; Based on the initial feature extraction sub-model and the sample speech features, performing keyword detection on the speech sample data to obtain predicted keywords; Calculating feature extraction loss data based on the predicted keywords and the key frame markers; Based on the initial adversarial classification sub-model and the sample speech features, the type of the speech sample data is identified to obtain a speech prediction category, wherein the speech prediction category is used to indicate that the speech sample data is the real speech data or the synthesized speech data; Calculating adversarial classification loss data based on the speech prediction category and the speech label; Based on the feature extraction loss data and the adversarial classification loss data, performing gradient reversal on the initial keyword detection model to obtain a target keyword detection model; Based on the target keyword detection model, keyword detection is performed on the target voice data to obtain the target keyword.

2. The method according to claim 1, characterized in that The step of performing gradient reversal on the initial keyword detection model based on the feature extraction loss data and the adversarial classification loss data to obtain a target keyword detection model includes: Calculating feature extraction gradient data based on the feature extraction loss data and preset model parameters, wherein the model parameters refer to parameters in the initial keyword detection model; Calculating adversarial classification gradient data based on the adversarial classification loss data and the model parameters; Performing data inversion on the adversarial classification gradient data to obtain inverted gradient data; Based on the feature extraction gradient data, the adversarial classification gradient data and the inversion gradient data, the initial keyword detection model is back-propagated to obtain the target keyword detection model.

3. The method according to claim 2, characterized in that The method of performing back propagation on the initial keyword detection model based on the feature extraction gradient data, the adversarial classification gradient data and the reverse gradient data to obtain the target keyword detection model includes: Based on the adversarial classification gradient data and the feature extraction gradient data, adjusting parameters of the initial adversarial classification sub-model to obtain a target adversarial classification sub-model; Based on the feature extraction gradient data and the inverted gradient data, adjusting parameters of the initial feature extraction sub-model to obtain a target feature extraction sub-model; The target feature extraction sub-model and the target adversarial classification sub-model are concatenated to obtain the target keyword detection model.

4. The method according to claim 1, characterized in that: The acquiring of speech sample data including speech tags and key frame tags includes: Get recorded voice data; Performing data identification on the recorded voice data to obtain the real voice data, wherein the real voice data includes a real label and a real key frame label, the real label is used to indicate that the artificially recorded data is the real voice data, and the real key frame label is used to identify the voice frame in the real voice data; Performing voice simulation on the real voice data to obtain simulated voice data; Performing data identification on the simulated speech data to obtain the synthesized speech data, wherein the synthesized speech data includes a synthesis tag and a synthesis key frame tag, the synthesis tag is used to indicate that the simulated speech data is the synthesized speech data, and the synthesis key frame tag is used to identify a speech frame in the synthesized speech data; The real voice data and the synthesized voice data are merged to obtain the voice sample data.

5. The method according to claim 4, characterized in that The performing voice simulation on the real voice data to obtain simulated voice data comprises: Performing voice transcription on the real voice data to obtain voice text data; Performing data enhancement on the speech text data to obtain enhanced text data; Speech synthesis is performed based on the enhanced text data to obtain the simulated speech data.

6. The method according to any one of claims 1 to 5, characterized in that: The step of performing keyword detection on the target voice data based on the target keyword detection model to obtain the target keyword includes: Based on the target keyword detection model, feature extraction is performed on the target speech data to obtain target speech features; Based on the target keyword detection model and the target speech features, performing context understanding on the target speech data to obtain context information; Based on the context information, feature classification is performed on the target speech feature to obtain a target feature category; Based on the target feature category, linearly mapping the target speech feature to obtain a keyword probability, wherein the keyword probability is used to indicate the probability that the target speech feature is a keyword; Based on the keyword probability, the target speech feature is screened to obtain the target keyword.

7. The method according to any one of claims 1 to 5, characterized in that: The method of performing type recognition on the speech sample data based on the initial adversarial classification sub-model and the sample speech features to obtain a speech prediction category includes: Based on the initial adversarial classification sub-model, feature recognition is performed on the sample speech features to obtain sample feature categories; Based on the sample feature category, data tracing is performed on the sample speech feature to obtain the speech prediction category.

8. A speech keyword detection device, characterized in that: The device comprises: A sample data acquisition module, used to acquire speech sample data including speech tags and key frame tags, wherein the speech sample data includes real speech data and synthesized speech data, the speech tags are used to indicate that the speech sample data is real speech data or synthesized speech data, and the key frame tags are used to identify speech frames in the speech sample data; An initial model acquisition module, used to acquire an initial keyword detection model, wherein the initial keyword detection model includes an initial feature extraction sub-model and an initial adversarial classification sub-model; A speech feature extraction module, used for performing feature extraction on the speech sample data based on the initial feature extraction sub-model to obtain sample speech features; A keyword prediction module, configured to perform keyword detection on the speech sample data based on the initial feature extraction sub-model and the sample speech features to obtain predicted keywords; A first loss calculation module, configured to calculate feature extraction loss data based on the predicted keywords and the key frame markers; A speech data classification module, used for performing type recognition on the speech sample data based on the initial adversarial classification sub-model and the sample speech features to obtain a speech prediction category, wherein the speech prediction category is used to indicate that the speech sample data is the real speech data or the synthesized speech data; A second loss calculation module, used for calculating adversarial classification loss data based on the speech prediction category and the speech label; A model parameter adjustment module is used to perform gradient inversion on the initial keyword detection model based on the feature extraction loss data and the adversarial classification loss data to obtain a target keyword detection model. The keyword detection module is used to perform keyword detection on the target speech data based on the target keyword detection model to obtain the target keyword.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the voice keyword detection method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the voice keyword detection method according to any one of claims 1 to 7 is implemented.

Images