A control method for speech recognition and computer device

The speech recognition model calculates the similarity of the target voiceprint data of speech information, which solves the problem of low speech recognition accuracy in noisy environments, and improves the recognition accuracy and the accuracy of command execution.

CN112259097BActive Publication Date: 2024-12-03SHENZHEN KONKA ELECTRONIC TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011164807.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-10-27
Publication Date
2024-12-03
Estimated Expiration
2040-10-27

AI Technical Summary

Technical Problem

The existing voice recognition technology has low accuracy in identifying voice information in noisy environments, and cannot effectively execute voice commands, and the environment requirements are high.

Method used

The target voiceprint data of voice information is obtained through the speech recognition model, the similarity to the standard voiceprint data is calculated, the response operation is determined, and the speech recognition accuracy is improved in noisy environments.

Benefits of technology

The voice information of the preset user is recognized among multiple voice information, which improves the accuracy of identifying voice information in noisy environments and realizes the accuracy of executing preset voice commands.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112259097B_ABST
    Figure CN112259097B_ABST
Patent Text Reader

Abstract

The present invention provides a control method and computer device for speech recognition, the control method for speech recognition comprising: obtaining at least one speech message including a keyword, and inputting the obtained at least one speech message into a trained speech recognition model, determining target voiceprint data corresponding to each of the at least one speech message through the speech recognition model; determining target similarity according to all determined target voiceprint data and a preset standard voiceprint data set; determining a response operation corresponding to at least one speech message according to the target similarity, and executing the response operation. In the present invention, the standard voiceprint data is the voiceprint data corresponding to a preset user, and the target voiceprint data corresponding to the target similarity is closest to the standard voiceprint data, so that the voice message sent by the preset user can be identified from multiple voice messages, thereby improving the accuracy of identifying voice information in a noisy environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of speech recognition technology, and in particular to a speech recognition control method and computer device. Background Art

[0002] Speech recognition can be used to control machines, where the machine recognizes speech information to obtain instructions corresponding to the speech information and executes the instructions. For example, when controlling a television by voice, the television can recognize speech information to receive instructions corresponding to the speech information and respond to the instructions.

[0003] Current speech recognition usually recognizes keywords in speech information. Recognition of keywords requires a quiet environment, and the speech information to be recognized needs to have a high decibel level in order to accurately recognize the instructions corresponding to the speech information. In the prior art, if the environment is noisy, the machine cannot accurately recognize the speech information, resulting in the execution of wrong instructions. In other words, the existing speech recognition has high requirements for the environment in which speech is recognized, and the accuracy of speech recognition is low in a noisy environment.

[0004] Therefore, the prior art needs to be further improved. Summary of the invention

[0005] The present invention provides a control method and computer device for speech recognition. In the present invention, a speech recognition model is used to determine target voiceprint data corresponding to at least one speech information including a keyword obtained simultaneously, and a target similarity is determined. Then, among multiple speech information including a keyword obtained simultaneously, the speech information corresponding to the target similarity is responded to, thereby improving the accuracy of recognizing speech information in a noisy environment.

[0006] In a first aspect, an embodiment of the present invention provides a method for controlling speech recognition, comprising:

[0007] Acquire at least one voice information including a keyword, and input the acquired at least one voice information into a voice recognition model, and determine the target voiceprint data corresponding to each of the at least one voice information through the voice recognition model;

[0008] Determine the target similarity based on all the determined target voiceprint data and a preset standard voiceprint data set;

[0009] Determine a response operation corresponding to the at least one voice information according to the target similarity, and execute the response operation.

[0010] In a further implementation, the standard voiceprint data set includes a plurality of standard voiceprint data; and determining the target similarity according to all the determined target voiceprint data and the preset standard voiceprint data set specifically includes:

[0011] For each target voiceprint data, respectively calculate the reference similarity between the target voiceprint data and each standard voiceprint data;

[0012] The highest reference similarity among all the calculated reference similarities is taken as the target similarity.

[0013] In a further implementation, determining a response operation corresponding to the at least one voice information according to the target similarity and executing the response operation specifically includes:

[0014] When the target similarity satisfies a preset condition, obtaining a target keyword in the voice information corresponding to the target similarity;

[0015] A target instruction is determined according to the target keyword, and a response operation corresponding to the target instruction is executed.

[0016] In a further implementation, determining a response operation corresponding to the at least one voice information according to the target similarity and executing the response operation specifically includes:

[0017] When the target similarity does not meet the preset condition, obtaining the time interval from the last execution of the response operation;

[0018] If the time interval is shorter than the preset time length, the voice message will not be responded to.

[0019] In a further implementation, when the target similarity does not meet the preset condition, after obtaining the time interval from the last time the response operation was performed, the method further includes:

[0020] If the time interval is greater than the preset time length, a target keyword in the voice information corresponding to the target similarity is obtained, a target instruction is determined according to the target keyword, and a response operation corresponding to the target instruction is executed.

[0021] In a further implementation, the preset condition is: the target similarity is greater than or equal to a preset threshold.

[0022] In a further implementation, the trained speech recognition model is obtained by training a preset network model based on a preset training speech set, wherein the training speech set includes multiple training speech and a classification label corresponding to each training speech.

[0023] In a further implementation, the training process of the speech recognition model specifically includes:

[0024] Inputting a training speech in the training speech set into a preset network model, and determining a prediction label corresponding to the training speech through the preset network model, wherein the training speech set includes a plurality of training speech and a classification label corresponding to each training speech;

[0025] Modify the model parameters of the preset network model according to the classification label corresponding to the training speech and the prediction label corresponding to the training speech, and continue to execute the step of inputting the training speech in the training speech set into the preset network model until the preset training conditions of the preset network model are met to obtain a speech recognition model.

[0026] In a second aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0027] Acquire at least one voice information including a keyword, and input the acquired at least one voice information into a voice recognition model, and determine the target voiceprint data corresponding to each of the at least one voice information through the voice recognition model;

[0028] Determine the target similarity based on all the determined target voiceprint data and a preset standard voiceprint data set;

[0029] Determine a response operation corresponding to the at least one voice information according to the target similarity, and execute the response operation.

[0030] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0031] Acquire at least one voice information including a keyword, and input the acquired at least one voice information into a voice recognition model, and determine the target voiceprint data corresponding to each of the at least one voice information through the voice recognition model;

[0032] Determine the target similarity based on all the determined target voiceprint data and a preset standard voiceprint data set;

[0033] Determine a response operation corresponding to the at least one voice information according to the target similarity, and execute the response operation.

[0034] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0035] In an embodiment of the present invention, at least one voice message including a keyword is obtained, and the obtained at least one voice message is input into a trained voice recognition model, and the target voiceprint data corresponding to each of the at least one voice message is determined by the voice recognition model; the target similarity is determined according to all the extracted target voiceprint data and a preset standard voiceprint data set; the response operation corresponding to the at least one voice message is determined according to the target similarity, and the response operation is executed. In the present invention, the standard voiceprint data is the voiceprint data corresponding to the preset user, and the target similarity is determined by the target voiceprint data corresponding to each voice message, and the target voiceprint data corresponding to the target similarity is closest to the standard voiceprint data. In this way, the voice message sent by the preset user can be identified from multiple voice messages, and then the response operation corresponding to the voice message sent by the preset user can be executed. The present invention responds to the voice message corresponding to the target similarity in multiple voice messages including keywords obtained at the same time, thereby improving the accuracy of recognizing voice information in a noisy environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 A flow chart of a method for controlling speech recognition according to an embodiment of the present invention;

[0038] Figure 2 A flowchart of a method for controlling speech recognition when specifically implemented in an embodiment of the present invention;

[0039] Figure 3 1 is a diagram showing the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0041] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0042] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as herein.

[0043] After research, the inventor found that current speech recognition usually recognizes keywords in speech information. Recognition of keywords requires a quiet environment, and the voice information to be recognized needs to have a high decibel level in order to accurately recognize the instructions corresponding to the voice information. In the prior art, if the environment is noisy, the machine cannot accurately recognize the voice information, resulting in the execution of wrong instructions. In other words, the existing speech recognition has high requirements for the environment in which speech is recognized, and the accuracy of speech recognition is low in a noisy environment.

[0044] In order to solve the above problems, in an embodiment of the present invention, at least one voice message including a keyword is obtained, and the obtained at least one voice message is input into a trained voice recognition model, and the target voiceprint data corresponding to each of the at least one voice message is determined by the voice recognition model; the target similarity is determined according to all the extracted target voiceprint data and a preset standard voiceprint data set; the response operation corresponding to the at least one voice message is determined according to the target similarity, and the response operation is executed. In the present invention, the standard voiceprint data is the voiceprint data corresponding to the preset user, and the target similarity is determined by the target voiceprint data corresponding to each voice message, and the target voiceprint data corresponding to the target similarity is closest to the standard voiceprint data. In this way, the voice message sent by the preset user can be identified from multiple voice messages, and then the response operation corresponding to the voice message sent by the preset user can be executed. The present invention responds to the voice message corresponding to the target similarity in the multiple voice messages including keywords obtained at the same time, thereby improving the accuracy of recognizing voice information in a noisy environment.

[0045] The invention is further described below by describing embodiments in conjunction with the accompanying drawings.

[0046] See also Figure 1 , this embodiment provides a method for controlling speech recognition, including:

[0047] S1. Acquire at least one voice information including a keyword, and input the acquired at least one voice information into a trained voice recognition model, and determine the target voiceprint data corresponding to each of the at least one voice information through the voice recognition model.

[0048] In an embodiment of the present invention, at least one voice message is obtained, and each voice message includes a keyword. The method of obtaining at least one voice message including a keyword may be: obtaining voice data, recognizing the voice data to obtain at least one keyword, and then determining at least one voice message including a keyword. The number of keywords in the voice data is the same as the number of voice messages including keywords. The voice data can be collected by a microphone, and the microphone for collecting voice data is connected to the terminal for communication, or a microphone is provided on the terminal.

[0049] In the embodiment of the present invention, it is possible to obtain voice data at fixed intervals, determine keywords in the obtained voice data, and then determine at least one voice information including the keyword. For example, if the voice data includes 3 keywords, then 3 voice information including the keywords are obtained.

[0050] In an embodiment of the present invention, the speech recognition model is obtained by training a preset network model based on a preset training speech set, wherein the training speech set includes multiple training speech and a classification label corresponding to each training speech. The training process of the speech recognition model will be described in detail later.

[0051] S2. Determine the target similarity based on all the determined target voiceprint data and a preset standard voiceprint data set.

[0052] In the embodiment of the present invention, the standard voiceprint data set includes several standard voiceprint data, and the standard voiceprint data can be the standard voiceprint data corresponding to several preset users. For example, a terminal can predetermine the standard voiceprint data corresponding to two users. Specifically, the voices of the two users are recorded into the terminal, and the standard voiceprint data corresponding to the two users are obtained through the speech recognition model to obtain the standard voiceprint data set. In actual use, the user can be prompted to say a specified text to collect the standard voiceprint data corresponding to the preset user.

[0053] In the embodiment of the present invention, after voice information including keywords is obtained, the target voiceprint data corresponding to each voice information is determined, and the reference similarity between each target voiceprint data and the standard voiceprint data is calculated to determine the target similarity.

[0054] Specifically, step S2 includes:

[0055] S21. For each target voiceprint data, respectively calculate the reference similarity between the target voiceprint data and each standard voiceprint data.

[0056] In the embodiment of the present invention, the target voiceprint data and the standard voiceprint data are both in vector form, and the reference similarity between the target voiceprint data and the standard voiceprint data can be calculated, and the reference similarity is in numerical form.

[0057] For example, there are two target voiceprint data, namely m1 and m2, and the standard voiceprint data set includes three standard voiceprint data, namely b1, b2 and b3; the reference similarity c11 between m1 and b1 is calculated, the reference similarity c12 between m1 and b2 is calculated, the reference similarity c13 between m1 and b3 is calculated, the reference similarity c21 between m2 and b1 is calculated, the reference similarity c22 between m2 and b2 is calculated, and the reference similarity c23 between m2 and b3 is calculated.

[0058] S22: taking the highest reference similarity among all the calculated reference similarities as the target similarity.

[0059] In the embodiment of the present invention, the reference similarity is in numerical form, and thus, the larger the numerical value corresponding to the reference similarity is, the larger the reference similarity is. The highest reference similarity among all the reference similarities is used as the target similarity.

[0060] When there are multiple highest reference similarities, for the multiple reference similarities that are tied for the highest, any one of the reference similarities can be selected as the target similarity, or the priorities corresponding to each of the multiple reference similarities are obtained (the priority is the preset priority of the user), and the reference similarity with the highest priority among the multiple reference similarities is used as the target similarity.

[0061] S3. Determine a response operation corresponding to the at least one voice information according to the target similarity, and execute the response operation.

[0062] In the embodiment of the present invention, there is only one target similarity. In the at least one voice information, the voice information corresponding to the target similarity is determined to be the target voice information, and the keyword included in the target voice information is the target keyword.

[0063] Specifically, step S3 includes:

[0064] S31A: When the target similarity satisfies a preset condition, a target keyword in the voice information corresponding to the target similarity is obtained.

[0065] In the embodiment of the present invention, the preset condition is that the target similarity is greater than or equal to a preset threshold, and the preset threshold can be set to 0.8. When the target similarity meets the preset condition, in the at least one voice information, the voice information corresponding to the target similarity is determined to be the target voice information, and the keyword included in the target voice information is the target keyword.

[0066] S32A: Determine a target instruction according to the target keyword, and execute a response operation corresponding to the target instruction.

[0067] In the embodiment of the present invention, the target keyword and the target instruction have a corresponding relationship. After determining the target keyword, the target instruction corresponding to the target keyword can be obtained, and then the response operation corresponding to the target instruction is executed. For example, if the target keyword is "play music", then the target instruction is an instruction for playing music. The terminal starts playing music according to the target instruction, that is, the response operation corresponding to the target instruction is executed.

[0068] In one example, step S3 includes:

[0069] S31B: When the target similarity does not meet the preset condition, obtain the time interval from the last time a response operation was performed.

[0070] S32B: If the time interval is less than the preset duration, do not respond to the voice message.

[0071] In an embodiment of the present invention, when the target similarity is less than a preset condition, it means that the target similarity is less than a preset threshold, that is, the target voiceprint data corresponding to the target similarity is not highly similar to the standard voiceprint data. It can be understood that at least one voice message is not sent by a preset user.

[0072] In the embodiment of the present invention, the so-called time interval refers to the time between the current (acquiring at least one voice message including a keyword) and the last time the response operation was performed. The preset time length can be set to 2 seconds or 5 seconds. When the time interval is less than the preset time length, the voice message is not responded.

[0073] S33B: If the time interval is greater than the preset time length, a target keyword in the voice information corresponding to the target similarity is obtained, a target instruction is determined according to the target keyword, and a response operation corresponding to the target instruction is executed.

[0074] In the embodiment of the present invention, if the time interval is greater than the preset time length, even though the voice information corresponding to the target similarity is not sent by the preset user, a response can be made.

[0075] Next, we will introduce the training process of the speech recognition model.

[0076] The speech recognition model is obtained by training a preset network model based on a preset training speech set, wherein the training speech set includes multiple training speech and a classification label corresponding to each training speech.

[0077] The training process of the speech recognition model includes:

[0078] The training speech in the training speech set is input into a preset network model, and the prediction label corresponding to the training speech is determined by the preset network model, wherein the training speech set includes multiple training speech and a classification label corresponding to each training speech; the model parameters of the preset network model are modified according to the classification label corresponding to the training speech and the prediction label corresponding to the training speech, and the step of inputting the training speech in the training speech set into the preset network model is continued until the preset training conditions of the preset network model are met to obtain a speech recognition model.

[0079] The speaker can be determined by voiceprint data, and the speaker's identity can be identified by sound. In an embodiment of the present invention, the preset network model first performs time-frequency analysis on the training speech, such as short-time Fourier transform, to obtain a training time-frequency graph corresponding to the training speech, extracts features of the training time-frequency graph through the preset network model, and classifies the extracted features to obtain a predicted label.

[0080] In the embodiment of the present invention, the training speech has a corresponding real classification label, and the real classification label is used to reflect the real speaker corresponding to the training speech. The classification label is represented by a vector, and the element corresponding to the speaker in the vector is 1, and the other elements are 0. The predicted label is represented by a vector, and each element in the predicted label is used to reflect the probability of the training speech predicted by the preset network model corresponding to each classification pair in the training speech set.

[0081] For example, the training speech set includes: training speech 1 corresponding to category 1, training speech 2 corresponding to category 2, and training speech 3 corresponding to category 3; that is, the classification label corresponding to training speech 2 is category 2, and the classification label is represented by a vector, in which the element corresponding to the classification is 1, and the remaining elements are 0. The classification label corresponding to training speech 2 can be (0,1,0). Among them, category 1 actually reflects a speaker, and the speaker corresponding to category 2 is different from the speaker corresponding to category 1. The predicted label can be (0.6, 0.2, 0.2), indicating that the preset network model believes that the probability that the classification corresponding to the input training speech is category 1 is the highest. In this example, for the sake of ease of explanation, only three training speech are simply given. In actual training, there can be thousands or tens of thousands of training speech in the training speech set.

[0082] In an embodiment of the present invention, a loss function value is calculated according to the prediction label and the classification label, the model parameters of the preset model are modified based on the calculated loss function value, and the step of inputting the training speech in the training speech set into the preset network model is continued until the preset training condition is met to obtain a speech recognition model. Cross entropy can be selected as the loss function.

[0083] In an embodiment of the invention, the preset training conditions include that the loss function value meets the preset requirements or the number of trainings reaches the preset number. The preset requirements can be determined according to the accuracy of the speech recognition model, which is not described in detail here. The preset number can be the maximum number of trainings of the preset network model, for example, 4000 times. Therefore, after the loss function value is calculated, it is determined whether the loss function value meets the preset requirements; if the loss function value meets the preset requirements, the training is terminated; if the loss function value does not meet the preset requirements, it is determined whether the number of trainings of the preset network model reaches the predicted number. If the preset number is not reached, the network parameters of the preset network model are corrected according to the loss function value; if the preset number is reached, the training is terminated. In this way, by judging whether the training of the preset network model is terminated by the loss function value and the number of trainings, it is possible to avoid the training of the preset network model entering an infinite loop due to the failure of the loss function value to meet the preset requirements.

[0084] Further, since the network parameters of the preset network model are modified when the training situation of the preset network model does not meet the preset training conditions, after the network parameters of the preset network model are corrected according to the loss function value, it is necessary to continue to train the preset network model, that is, continue to execute the step of inputting the training speech in the training speech set into the preset network model. Among them, continuing to execute the training speech in the training speech set into the preset network model can be a training speech that has never been input into the preset network model as an input item. For example, all training speech in the training data have a unique identifier (for example, a number), and the identifier of the training speech input into the preset network model for the first training is different from the identifier of the training speech input into the preset network model for the second training. Of course, in actual applications, since the number of training speech in the training speech set is limited, in order to improve the training effect of the speech recognition model, the training speech in the training speech set can be sequentially input into the preset network model to train the preset network model. When all the training speech in the training speech set are input into the preset network model, the operation of inputting the training speech in the training speech set into the preset network model can be continued. In this embodiment, the specific implementation method of "continuing to execute the step of inputting the training speech in the training speech set into the preset network model" is not limited.

[0085] In an embodiment of the present invention, after the preset training conditions are met, a speech recognition model is obtained, the parameters of the speech recognition model can be fixed, and the speech recognition model is used on the terminal; or, after the preset training conditions are met, a speech recognition model is obtained, the parameters of the speech recognition model are not fixed, and the speech recognition model continues to be trained during use.

[0086] Specifically, the terminal obtains voice information including keywords, recognizes the voice information through the voice recognition model, and then obtains the voice recognition result, which can be displayed. The voice recognition result can be: the user is A, the user selects the real user, and the model parameters of the voice recognition model are continuously adjusted according to the real user and the voice recognition result. The model parameters of the voice recognition model can be continuously optimized and adjusted during use to obtain a voice recognition model with better classification accuracy. For example, user A issues an instruction, and the recognition result of the voice recognition model is user B. The terminal pops up a prompt box, and the prompt box gives options for user A to select the real speaker. User A selects the option "user A" in the prompt box, then the loss function value is calculated according to the "user B" predicted by the voice recognition model and the real speaker "user A", and the model parameters of the voice recognition model are continuously modified.

[0087] In the specific implementation, see Figure 2 , a speech recognition control method includes the following process.

[0088] When obtaining voice information including keywords, the voice information including keywords is input into the voice recognition model to obtain the target voiceprint data. When there are multiple voice information including keywords, the target voiceprint data corresponding to each of the multiple voice information can be obtained through the voice recognition model. The target similarity can be determined based on the standard voiceprint data set and all the target voiceprint data obtained. When the target similarity meets the preset conditions, it can be considered that the target keyword corresponding to the target similarity is issued by the preset user, and the target keyword corresponding to the target similarity is directly responded to; when the target similarity does not meet the preset conditions, it means that the target keyword corresponding to the target similarity is not issued by the preset user, then the time interval from the last response operation is obtained. If the time interval is greater than the preset duration, the target keyword is responded to. That is to say, if no preset user issues an instruction in a short period of time, the terminal can also respond to the target keyword not given by the user.

[0089] In an embodiment of the present invention, at least one voice message including a keyword is obtained, and the obtained at least one voice message is input into a trained voice recognition model, and the target voiceprint data corresponding to each of the at least one voice message is determined by the voice recognition model; the target similarity is determined according to all the extracted target voiceprint data and a preset standard voiceprint data set; the response operation corresponding to the at least one voice message is determined according to the target similarity, and the response operation is executed. In the present invention, the standard voiceprint data is the voiceprint data corresponding to the preset user, and the target similarity is determined by the target voiceprint data corresponding to each voice message, and the target voiceprint data corresponding to the target similarity is closest to the standard voiceprint data. In this way, the voice message sent by the preset user can be identified from multiple voice messages, and then the response operation corresponding to the voice message sent by the preset user can be executed. The present invention responds to the voice message corresponding to the target similarity in multiple voice messages including keywords obtained at the same time, thereby improving the accuracy of recognizing voice information in a noisy environment.

[0090] In one embodiment, the present invention provides a computer device, which may be a terminal, and the internal structure is as follows: Figure 3As shown. The computer device includes a processor, a memory, a network model interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network model interface of the computer device is used to communicate with an external terminal through a network model connection. When the computer program is executed by the processor, a control method for speech recognition is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0091] Those skilled in the art will understand that Figure 3 What is shown is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0092] An embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0093] Acquire at least one voice information including a keyword, and input the acquired at least one voice information into a voice recognition model, and determine the target voiceprint data corresponding to each of the at least one voice information through the voice recognition model;

[0094] Determine the target similarity based on all the determined target voiceprint data and a preset standard voiceprint data set;

[0095] Determine a response operation corresponding to the at least one voice information according to the target similarity, and execute the response operation.

[0096] The embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the following steps are implemented:

[0097] Acquire at least one voice information including a keyword, and input the acquired at least one voice information into a voice recognition model, and determine the target voiceprint data corresponding to each of the at least one voice information through the voice recognition model;

[0098] Determine the target similarity based on all the determined target voiceprint data and a preset standard voiceprint data set;

[0099] Determine a response operation corresponding to the at least one voice information according to the target similarity, and execute the response operation.

[0100] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0101] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for controlling speech recognition, It is characterized in that include: Acquire at least one voice information including a keyword, and input the acquired at least one voice information into a voice recognition model, and determine the target voiceprint data corresponding to each of the at least one voice information through the voice recognition model; Determine the target similarity based on all the determined target voiceprint data and a preset standard voiceprint data set; Determine a response operation corresponding to the at least one voice information according to the target similarity, and execute the response operation; The standard voiceprint data set includes a plurality of standard voiceprint data; The determining of the target similarity based on all the determined target voiceprint data and the preset standard voiceprint data set specifically includes: For each target voiceprint data, respectively calculate the reference similarity between the target voiceprint data and each standard voiceprint data; The highest reference similarity among all the calculated reference similarities is taken as the target similarity; The number of keywords in the voice data is the same as the number of voice messages containing the keywords; The step of determining a response operation corresponding to the at least one voice information according to the target similarity and executing the response operation specifically includes: When the target similarity does not meet the preset condition, obtaining the time interval from the last execution of the response operation; If the time interval is less than the preset time length, the voice message is not responded to; When the target similarity does not meet the preset condition, after obtaining the time interval from the last time the response operation was performed, the method further includes: If the time interval is greater than the preset time length, a target keyword in the voice information corresponding to the target similarity is obtained, a target instruction is determined according to the target keyword, and a response operation corresponding to the target instruction is executed; When there are multiple highest reference similarities, the priorities corresponding to the multiple highest reference similarities are obtained, and the reference similarity with the highest priority among the multiple highest reference similarities is used as the target similarity.

2. The method for controlling speech recognition according to claim 1, It is characterized in that The step of determining a response operation corresponding to the at least one voice information according to the target similarity and executing the response operation specifically includes: When the target similarity meets the preset condition, obtaining the target keyword in the voice information corresponding to the target similarity; A target instruction is determined according to the target keyword, and a response operation corresponding to the target instruction is executed.

3. The method for controlling speech recognition according to claim 2, It is characterized in that The preset condition is: the target similarity is greater than or equal to a preset threshold.

4. The method for controlling speech recognition according to claim 1, It is characterized in that The trained speech recognition model is obtained by training a preset network model based on a preset training speech set, wherein the training speech set includes multiple training speech and a classification label corresponding to each training speech.

5. The method for controlling speech recognition according to claim 4, It is characterized in that The training process of the speech recognition model specifically includes: Inputting a training speech in the training speech set into a preset network model, and determining a prediction label corresponding to the training speech through the preset network model, wherein the training speech set includes a plurality of training speech and a classification label corresponding to each training speech; Modify the model parameters of the preset network model according to the classification label corresponding to the training speech and the prediction label corresponding to the training speech, and continue to execute the step of inputting the training speech in the training speech set into the preset network model until the preset training conditions of the preset network model are met to obtain a speech recognition model.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, It is characterized in that When the processor executes the computer program, the steps in the method for controlling speech recognition according to any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps in the method for controlling speech recognition according to any one of claims 1 to 5 are implemented.