A speech recognition method, apparatus, electronic device, and storage medium

Abstract

The application provides a speech recognition method and device, electronic equipment and a storage medium, wherein the speech recognition method comprises: extracting a voiceprint feature from a speech to be recognized; determining whether a first speaker corresponding to the speech has a speech transcription right according to the voiceprint feature; if the first speaker corresponding to the speech has the speech transcription right, performing speech recognition processing on the speech to obtain a speech text of the first speaker. The scheme can flexibly determine whether to perform speech recognition processing on the speech of the speaker, thereby improving the flexibility of speech recognition, and being beneficial to alleviating the problems of tight speech recognition resources and waste of recognition resources.

Description

Technical Field

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

[0002] With the rapid development of artificial intelligence and the widespread application of AI technology, AI speech recognition technology has been involved in all aspects of people's lives.

[0003] When the existing speech recognition systems are turned on, they recognize the speech of different speakers and generate corresponding speech text. However, when faced with different speech recognition needs, indiscriminately recognizing the speech of different speakers cannot meet the actual needs of different speech recognition scenarios, and it can also lead to problems such as limited recognition resources and waste of recognition resources. Therefore, the flexibility of existing speech recognition systems cannot meet actual speech recognition needs.

[0004] Therefore, how to improve the flexibility of speech recognition has become a technical problem that technical personnel in this field urgently need to solve. Summary of the Invention

[0005] In view of this, the present application provides a speech recognition method, device, electronic device and storage medium to improve the flexibility of speech recognition.

[0006] According to a first aspect of an embodiment of the present application, a speech recognition method is provided, comprising:

[0007] Extracting voiceprint features from the speech to be recognized;

[0008] Determining, based on the voiceprint features, whether the first speaker corresponding to the speech has voice transcription authority;

[0009] If the first speaker corresponding to the speech voice has the speech transcription permission, speech recognition processing is performed on the speech voice to obtain the speech text of the first speaker.

[0010] In an optional implementation manner of the present application, determining, based on the voiceprint feature, whether the first speaker corresponding to the speech has voice transcription permission includes:

[0011] determining the identity information of the first speaker from a voiceprint database according to the voiceprint feature;

[0012] According to a preset voice transcription permission table and the identity information of the first speaker, it is determined whether the first speaker has the voice transcription permission; wherein the preset voice transcription permission table records the voice transcription permissions of multiple speakers.

[0013] In an optional embodiment of the present application, the voiceprint database is obtained by:

[0014] Obtaining preset multiple speaker voice sample information;

[0015] respectively extracting voiceprint features from the speech sample information of the plurality of speakers;

[0016] The voiceprint database is constructed based on the correspondence between the voiceprint features in the voice sample information and the identity information of the multiple speakers.

[0017] In an optional embodiment of the present application, the present invention further includes:

[0018] Determining text access identification information of the first speaker, where the text access identification information includes permission information for other persons to access the speech text of the first speaker;

[0019] Based on the text review identification information of the first speaker, review authority for the speech text of the first speaker is set.

[0020] In an optional embodiment of the present application, the present invention further includes:

[0021] Obtaining a speech text of the second speaker, wherein the first speaker and the second speaker have the same access rights to the speech texts;

[0022] Determining text similarity between a speech text of the first speaker and a speech text of the second speaker;

[0023] When the text similarity is greater than a preset similarity threshold, the speech text of the first speaker and the speech text of the second speaker are merged to obtain a merged speech text.

[0024] In an optional implementation manner of the present application, performing speech recognition processing on the speech to obtain the speech text of the first speaker includes:

[0025] Retrieving a personalized transcription model corresponding to the first speaker from a preset personalized transcription resource library based on the voiceprint features of the first speaker; the personalized transcription model is obtained by performing recognition training on speech with specific speech characteristics, and the speech of the first speaker has the specific speech characteristics;

[0026] If a personalized transcription model corresponding to the first speaker is retrieved from a preset personalized transcription resource library, speech recognition processing is performed on the speech of the first speaker using the personalized transcription model to obtain a speech text of the first speaker.

[0027] In an optional implementation of the present application, the personalized transcription model corresponding to the first speaker is obtained by performing speech recognition training on the speech of the first speaker.

[0028] In an optional embodiment of the present application, the present invention further includes:

[0029] Determining whether an adjustment instruction for the voice transcription authority of the first speaker is obtained;

[0030] If an adjustment instruction for the voice transcription authority of the first speaker is obtained, the voice transcription authority of the first speaker is adjusted according to the adjustment instruction.

[0031] According to a second aspect of an embodiment of the present application, a speech recognition device is provided, including:

[0032] An extraction unit, configured to extract voiceprint features from the speech to be recognized;

[0033] a judgment unit, configured to judge, based on the voiceprint feature, whether the first speaker corresponding to the speech has the voice transcription permission;

[0034] The recognition unit is configured to perform speech recognition processing on the speech voice to obtain a speech text of the first speaker if the first speaker corresponding to the speech voice has speech transcription authority.

[0035] According to a third aspect of the embodiments of the present application, there is provided an electronic device, including:

[0036] processor;

[0037] a memory for storing instructions executable by the processor;

[0038] The processor is configured to execute any one of the above-mentioned speech recognition methods by running instructions in the memory.

[0039] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, including: the storage medium stores a computer program, and when the computer program is executed by a processor, any one of the above-mentioned speech recognition methods is executed.

[0040] Compared with the prior art, this application has the following advantages:

[0041] The speech recognition method, device, electronic device and storage medium provided in this application determine whether the first speaker corresponding to the speech to be recognized has the speech transcription authority through the voiceprint features in the speech to be recognized, and then perform speech recognition processing on the speech based on the speech transcription authority of the first speaker. Before performing speech recognition on the speaker's speech, this method can automatically determine whether the speaker has the speech transcription authority, and only perform speech recognition processing on the speaker's speech when it is determined that the speaker has the speech transcription authority. This solution can flexibly decide whether to perform speech recognition processing on the speaker's speech, thereby improving the flexibility of speech recognition and helping to alleviate the problems of tight speech recognition resources and waste of recognition resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0043] Figure 1 A schematic diagram of an application scenario of the speech recognition method provided in one embodiment of the present application;

[0044] Figure 2 A flow chart of a speech recognition method provided in another embodiment of the present application;

[0045] Figure 3 A training flow chart of a personalized transcription model provided in another embodiment of the present application;

[0046] Figure 4 A schematic diagram of the network structure of a speech text regularization network provided in another embodiment of the present application;

[0047] Figure 5 A schematic diagram of the structure of a speech recognition device provided in another embodiment of the present application;

[0048] Figure 6 A schematic diagram of the structure of an electronic device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0050] With the rapid development of artificial intelligence and the widespread application of AI technology, AI speech recognition technology has been involved in all aspects of people's lives.

[0051] When the existing speech recognition systems are turned on, they recognize the speech of different speakers and generate corresponding speech text. However, when faced with different speech recognition needs, indiscriminately recognizing the speech of different speakers cannot meet the actual needs of different speech recognition scenarios, and it can also lead to problems such as limited recognition resources and waste of recognition resources. Therefore, the flexibility of existing speech recognition systems cannot meet actual speech recognition needs.

[0052] Therefore, how to improve the flexibility of speech recognition has become a technical problem that technical personnel in this field urgently need to solve.

[0053] In order to solve the above technical problems, the present application provides a speech recognition method, device, electronic device and storage medium, which will be described in detail one by one in the following embodiments.

[0054] Exemplary Implementation Environment

[0055] First, in order to facilitate understanding of the specific application scenario of the speech recognition method provided in this application, the speech recognition method is introduced below in conjunction with an application scenario of the speech recognition method.

[0056] Please refer to Figure 1 , Figure 1 A schematic diagram of an application scenario of the speech recognition method provided in one embodiment of the present application.

[0057] like Figure 1 As shown, Figure 1 The illustrated speech recognition scenario is specifically about performing speech recognition on a person when the person attends a meeting and speaks.

[0058] Figure 1 The system includes: a speaker 101 and a speech recognition device 102.

[0059] The speaker 101 may be understood as a person who participates in the meeting and is speaking in the meeting.

[0060] The speech recognition device 102 can be understood as a speech recognition device for performing speech recognition on a speaker. In an optional embodiment of the present application, the speech recognition device 102 can be a hardware device such as a voice recorder, a voice recorder, a mobile phone, a computer, or a smart terminal.

[0061] In the process of the speech recognition device 102 recognizing the speaker, the following steps are performed: Figure 1Steps S110 to S120 shown:

[0062] Step S110, start;

[0063] Step S111, obtaining the voiceprint features of the spoken speech;

[0064] The voiceprint features of the speech can be obtained based on the speech information of the conference speaker when speaking in the conference;

[0065] Step S112, determining the speaker from the voiceprint database based on the voiceprint feature;

[0066] The voiceprint database stores the voiceprint features of multiple speakers participating in the meeting, as well as the corresponding relationship between the speaker's identity information and the voiceprint features. Therefore, the speaker who is currently speaking can be identified based on the corresponding relationship between the speaker's identity information and the voiceprint features recorded in the voiceprint database.

[0067] Step S113, determining whether the speaker has voice transcription permission;

[0068] Among them, the present application can pre-determine the speech transcription authority of multiple speakers, and store the speech transcription authority of the speakers and the identity information of the speakers in a pre-prepared speech transcription authority table, so that the speech recognition device 102 can directly call the speech transcription authority table, and then determine whether the person has the speech transcription authority based on the identity information of the speaker.

[0069] Before performing voice recognition on the speaker's speech, this application can automatically determine whether the speaker has the voice transcription permission. Only when it is determined that the speaker has the voice transcription permission will the speaker's speech be voice recognized, which improves the flexibility of voice recognition and helps alleviate the problems of tight voice recognition resources and waste of recognition resources.

[0070] If the speaker does not have the voice transcription permission, execute step S114;

[0071] Step S114, ending the transcription process;

[0072] If the speaker has the voice transcription permission, execute step S115;

[0073] Step S115: searching the personalized resource library to determine whether there is a personalized transcription model for the speaker;

[0074] In an optional embodiment of the present application, the personalized resource library stores a personalized transcription model and a universal transcription model; wherein the personalized transcription model and the universal transcription model can be understood as a convolutional neural network trained by machine learning.

[0075] Among them, the general transcription model can transcribe the speech of any speaker; different from the general transcription model, the personalized transcription model is a transcription model set for a designated speaker or a specific type of speaker. The model can be obtained through recognition training of speech with specific voice characteristics, wherein the designated speaker has the specific voice characteristics.

[0076] If there is no personalized transcription model, execute step S116;

[0077] Step S116, inputting the spoken speech into a universal transcription model;

[0078] If the personalized transcription model exists, execute step S117;

[0079] Step S117, inputting the spoken speech into a personalized transcription model;

[0080] This solution sets up a personalized transcription model for the speaker, so that the speech can be transcribed in combination with the speaker's specific voice characteristics, so as to improve the transcription accuracy of the speech in the subsequent stage.

[0081] Step S118, obtaining the speech text of the speaker;

[0082] Among them, the speech text of the speaker is the content output by the personalized transcription model or the universal transcription model. In the scenario embodiment provided in this application, the speech text can be understood as the meeting minutes of the speaker.

[0083] Step S119, setting access permissions for the speech text of the speaker;

[0084] The review authority may be understood as the authority for other people to review the speech text of the speaker.

[0085] In an optional embodiment of the present application, the access rights to the speech text of the speaker can be set according to the speaker's level and / or the confidentiality level of the speaker's speech content. By setting corresponding access rights for the speaker's speech text, the present application improves the confidentiality of the speech text, effectively avoids the leakage of sensitive information in the speech text, and reduces the difficulty of post-management of the speech text.

[0086] Step S120, end.

[0087] It can be understood that the above introduction to the embodiment of the scenario of this application is only to facilitate a better understanding of the speech recognition method provided by this application, and is not used to limit the application scenario of the speech recognition method. The speech recognition method can also be applied to other scenarios, for example, for recognizing call voices, recognizing recorded information, etc.

[0088] Exemplary Methods

[0089] In an exemplary embodiment of the present application, a speech recognition method is provided. The core of the method is to automatically determine whether the speaker has speech transcription permission before performing speech recognition on the speaker's speech. Only when it is determined that the speaker has speech transcription permission is speech recognition performed on the speaker's speech.

[0090] In an optional embodiment of the present application, the implementation subject of the speech recognition method can be various types of user terminals such as laptops, tablet computers, desktop computers, set-top boxes, mobile devices (for example, mobile phones, portable music players, personal digital assistants, dedicated messaging devices, game consoles), or a combination of any two or more of these data processing devices, or a server.

[0091] See also Figure 2 , Figure 2 A flow chart of a speech recognition method provided in another embodiment of the present application.

[0092] The method includes the following steps S201 to S203:

[0093] Step S201: extracting voiceprint features from the speech to be recognized.

[0094] The speech to be recognized can be understood as the speech of a speaker. For example, the speech information of the speaker currently speaking in an internal or public meeting of a company. The speech can be in any language, of any duration, and with any content.

[0095] In an optional embodiment of the present application, the speech to be recognized can be acquired by a pre-prepared sound sensor. For example, the sound sensor can be a mobile phone, computer, voice recorder, etc. of the speaker or other staff member, or a terminal device such as a recorder or microphone set up in the speaker's location specifically for speech acquisition.

[0096] In the field of speech recognition, the voiceprint can be understood as the spectrum of sound waves carrying speech information displayed by electroacoustic instruments. In real life, the voice of each person when speaking has its own characteristics, and the voice of a person has specificity and stability. Therefore, in the embodiments of the present application, the voiceprint features are used to determine the speaker corresponding to the speech to be recognized.

[0097] In an optional embodiment of the present application, the voiceprint features extracted from the speech to be recognized can be realized by a voiceprint feature extraction module dedicated to extracting voiceprint features; wherein the voiceprint feature extraction module is deployed with a corresponding voiceprint extraction algorithm or a machine learning model to realize the extraction of voiceprint features.

[0098] Step S202, according to the voiceprint features, judging whether the first speaker corresponding to the speech has the speech transcription right.

[0099] In an optional embodiment of the present application, the above step S202 can be realized by the following steps S1 and S2:

[0100] Step S1, according to the voiceprint features, determining the identity information of the first speaker from the voiceprint library.

[0101] The voiceprint library can be understood as a database for storing voiceprint features of different speakers and the corresponding relationship between voiceprint features and the identity information of the speaker.

[0102] In an optional embodiment of the present application, the voiceprint library can be obtained by the following way:

[0103] First, obtain the voice sample information of a plurality of preset speakers;

[0104] Second, extract the voiceprint features in the voice sample information of the plurality of speakers respectively;

[0105] Finally, based on the corresponding relationship between the voiceprint features in the voice sample information and the identity information of the plurality of speakers, the voiceprint library is constructed.

[0106] In practical application, the voice sample information of the plurality of speakers can be collected according to actual needs; for example, in a conference held in a certain company, the speech of a plurality of speakers participating in the conference needs to be recognized; in this scenario, the voice sample information of the plurality of preset speakers can be at least one piece of voice information of the participants collected by a voice recorder, a computer or other terminal device.

[0107] In another optional embodiment of the present application, in order to avoid the different speech styles of the speaker in different scenarios and improve the voiceprint quality of the voiceprint library in the current scenario, the voice sample information of the preset multiple speakers can be collected at the specific speaking location of the speaker, or the speaker's speech in the same or similar speaking scenario as the current speaking scenario can be collected based on platforms such as the Internet; for example, assuming that the speaker attends a formal meeting held regularly, the speaker's voice sample information can be one or more speech segments delivered by the speaker when he or she attended the formal meeting in the past.

[0108] Furthermore, the correspondence between the voiceprint features and the identity information of the multiple speakers can be obtained through a preset voiceprint registration module when the speaker registers the voiceprint information. For example, after obtaining the voice sample information of Zhang San, the voice sample information is first labeled with a name or number to indicate the identity information of the speaker (Zhang San) to whom the voice sample information belongs. Then, after extracting the voiceprint features of the voice sample information, the labeled name or number is matched with the voiceprint features to indicate the correspondence between the speaker's identity information and the voiceprint features.

[0109] Step S2: judging whether the first speaker has the speech transcription permission according to a preset speech transcription permission table and the identity information of the first speaker; wherein the preset speech transcription permission table records the speech transcription permissions of a plurality of speakers.

[0110] In the embodiment of the present application, the speech transcription permission table can be understood as a table pre-stored in the speech recognition system for storing whether different speakers have speech transcription permissions.

[0111] Please refer to Table 1, which is a speech transcription permission table provided in an embodiment of the present application.

[0112] Table 1

[0113] spokesman Voice transcription permission Zhang San 1 Li Si 0 Wang Wu 1 Zhao Liu 1

[0114] As shown in Table 1, the speech transcription permission table can be understood as a two-dimensional table that records the speech transcription permissions of different speakers. The two-dimensional table records the names of different speakers (such as Zhang San, Li Si, etc.) to represent the identity information of different speakers, and also records the speech transcription permissions of different speakers. Among them, 1 indicates that the speaker has the speech transcription permission, and 0 indicates that the speaker does not have the speech transcription permission. For example, in Table 1, Zhang San has the speech transcription permission, but Li Si does not have the speech transcription permission.

[0115] In an optional implementation of the present application, the speech transcription permission table can be dynamically adjusted according to actual conditions. For example, assuming that during the meeting, a new speaker needs to be added and the speaker needs to have speech transcription permissions, then at this time, the relevant staff can modify the speech transcription permission table through a computer or other terminal device to add a new speaker and set speech transcription permissions for him.

[0116] It is understandable that, because the aforementioned speech transcription permission table is dynamically adjustable, the speech transcription permission table can be a newly created permission table based on the speech recognition scenario, a historical permission table that meets actual speech recognition requirements, or a permission table modified based on actual needs based on a historical permission table. This application does not impose any restrictions on this.

[0117] In order to further improve the flexibility of speech transcription, the speech recognition method also includes: determining whether an adjustment instruction for the speech transcription authority of the first speaker is obtained; wherein, if an adjustment instruction for the speech transcription authority of the first speaker is obtained, the speech transcription authority of the first speaker is adjusted according to the adjustment instruction.

[0118] In an optional manner of the present application, the instruction to adjust the voice transcription authority may be issued by a relevant staff member who presides over the voice recognition device, such as a host or other staff member who presides over the meeting.

[0119] In an optional implementation of the present application, the transcription authority of the first speaker can be set according to at least one of the current speech topic of the first speaker, the historical speech transcription authority of the first speaker, and the actual demand information of the first speaker.

[0120] For example, if the first speaker mentions company confidential information during a meeting, and the speech is not suitable for speech recognition, the meeting host or relevant staff can temporarily disable the first speaker's speech transcription permission when the first speaker mentions such information.

[0121] For example, if the first speaker does not want his or her voice to be recognized, he or she can explain the situation to the relevant staff to modify his or her voice transcription permissions.

[0122] For another example, assuming that the first speaker has always had voice transcription permissions in previous meetings or voice recognition scenarios, then when the first speaker participates in the meeting, unless there are special requirements, the first speaker's voice transcription permissions can be kept enabled;

[0123] Step S203, if the first speaker corresponding to the speech voice has a voice transcription right, performing voice recognition processing on the speech voice to obtain the speech text of the first speaker.

[0124] In the embodiments of the present application, the purpose of performing voice recognition processing on the speech of the first speaker is to obtain the speech text of the first speaker to record the speech content of the first speaker. For example, in the scenario of conference voice recognition, the conference minutes of the first speaker can be generated based on the voice recognition processing on the speech voice of the first speaker.

[0125] In an optional embodiment of the present application, the above step S203 can be implemented through the following steps S3 and S4:

[0126] Step S3, retrieving a personalized transcription model corresponding to the first speaker from a preset personalized transcription resource library according to the voiceprint feature of the first speaker; the personalized transcription model is obtained by performing recognition training on a voice with specific voice characteristics, and the speech voice of the first speaker has the specific voice characteristics.

[0127] The preset personalized transcription resource library can be understood as a database for storing personalized transcription models and the correspondence between the personalized transcription models and speakers.

[0128] The personalized transcription model can be understood as a convolutional neural network. In the process of specific application, the personalized transcription model is obtained by training in the manner of machine learning (Machine Learning, ML). Machine learning (a multi-disciplinary subject, involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc.) is specialized in acquiring new knowledge or skills through training samples, reorganizing existing knowledge structure and constantly improving its own performance. Machine learning usually includes artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, etc., and is a branch of artificial intelligence (Artificial Intellingence, AI) technology.

[0129] In an optional embodiment of the present application, the personalized transcription model of the first speaker can be obtained by performing voice recognition training on a general transcription model with a voice with specific voice characteristics. The voice with specific voice characteristics used to train the personalized transcription model of the first speaker can be obtained by collecting through the Internet or by collecting the speech voice of the first speaker.

[0130] Furthermore, the specific language characteristics can be understood as unique voice characteristics of the first speaker, such as the first speaker's speaking speed and intonation. They can also refer to language characteristics of a particular type to which the first speaker belongs, such as the first speaker's dialect. It is understood that regardless of the language characteristics of the first speaker, they can be obtained by collecting the first speaker's speech information.

[0131] In an optional implementation of the present application, different personalized speech transcription models are set according to the language characteristics of different speakers to improve the transcription accuracy of the personalized transcription model.

[0132] For further information, please refer to Figure 3 , Figure 3 A training flowchart of a personalized transcription model provided in another embodiment of the present application.

[0133] Figure 3 It includes: a sample collection module 301, a model training module 302, and a model acquisition module 303.

[0134] The sample collection module 301 is used to obtain speech with specific speech characteristics of the first speaker as a training sample; wherein the training sample can be obtained through the Internet, or by recording the speech information of the first speaker.

[0135] In an optional embodiment of the present application, a speech voice in the same scene as the current speech scene of the first speaker and the speech text corresponding to the speech voice are used as training samples to further improve the compatibility and transcription accuracy of the first speaker and the personalized transcription model.

[0136] For example, assuming that the current speaking scenario of the first speaker is a formal meeting held regularly, speech clips of the first speaker in previous formal meetings and speech texts corresponding to the speech clips can be used as training samples.

[0137] The training samples are then sent to the model training module 302 , which fine-tunes the general transcription model so that the model acquisition module 303 obtains the personalized transcription model of the first speaker.

[0138] Step S4: If a personalized transcription model corresponding to the first speaker is retrieved from a preset personalized transcription resource library, speech recognition processing is performed on the speech of the first speaker using the personalized transcription model to obtain the speech text of the first speaker.

[0139] In an optional embodiment of the present application, the personalized transcription model adopts an end-to-end ED network structure based on an attention mechanism, wherein the ED network structure includes an encoder and a decoder, wherein the encoder is used to receive the speech of the first speaker in the input network structure and convert the speech into high-order features; the decoder is used to receive the high-order features and use the attention mechanism to convert the high-order features into a probability distribution of a character sequence, and based on this probability distribution, obtain the speech text of the first speaker.

[0140] In another optional embodiment of the present application, if the corresponding personalized transcription model of the first speaker is not retrieved from the preset personalized transcription resource library, the speech of the first speaker can be processed by speech recognition using a universal transcription model to obtain the speech text of the first speaker.

[0141] The universal transcription model is used to transcribe spoken speech to obtain speech text. Similar to the personalized transcription model, the universal transcription model is also a convolutional neural network. In actual application, the universal transcription model can use speech obtained through the internet or other means and the speech text corresponding to the speech as training samples, and obtain them through machine learning.

[0142] Furthermore, since the speech text generated by the personalized transcription model or the universal transcription model is basically consistent with the content of the speaker's speech, but considering that the speaker will inevitably use some colloquial content or some repetitive content during the speech, it is necessary to regularize the speaker's speech text to make the speech text more written while retaining the speaker's speech intention, thereby improving the quality of the speaker's speech text.

[0143] For example, the text of the speaker's speech may be "Well, today's meeting mainly revolves around three aspects, and the three aspects are discussed." After text regularization, the above speech text of the speaker can be "Today's meeting mainly revolves around three aspects."

[0144] In an optional implementation of the present application, a speech text regularization network may be used to process the speech text to obtain the regularized speech text.

[0145] Specifically, the speech text regularization network can be a pointer-generator network (Pointer-Generator Networks), please refer to Figure 4 , Figure 4A network structure diagram of a speech text normalization network is provided for another embodiment of the present application.

[0146] As shown in Figure 4 the speech text normalization network includes an input layer 401, a processing layer 402, and an output layer 403.

[0147] The input layer 401 is configured to input the speech text of the speaker.

[0148] The processing layer 402 is configured to perform normalization processing on the speech text, which includes but is not limited to filtering of emotional words in the speech text, normalization of numbers in the speech text, etc.

[0149] The output layer 403 is configured to output the normalized speech text.

[0150] Further, after obtaining the speech text of the first speaker through the above steps S201 to S203, in order to facilitate the sharing of the speech text, it is further necessary to set the access permission of the speech text, so as to realize the intelligent transmission of the text content and improve the efficiency of the conference content.

[0151] Further, the speech recognition method provided by the present application further includes the following steps S204 and S205:

[0152] Step S204: determining the text access identification information of the first speaker, wherein the text access identification information includes the permission information of other people accessing the speech text of the first speaker.

[0153] Step S205: setting the access permission of the speech text of the first speaker based on the text access identification information of the first speaker.

[0154] In an optional embodiment of the present application, the speech texts of different speakers can be mutually accessed according to needs. For example, in a conference, the speech content of the first speaker is necessary for other speakers to learn from, and then the other speakers can access the speech text of the first speaker according to the access permission.

[0155] The main function of the text access identification information is to identify the access permission information of other people accessing the speech text of the first speaker, and then to assign the corresponding access permission to the speech text of the first speaker according to the text access identification information of the first speaker.

[0156] Please refer to Table 2, which is an access permission table provided by an embodiment of the present application.

[0157] Table 2

[0158]

[0159] As shown in Table 2, the text reference identification information described in the above steps can be understood as the content recorded in the text level column and the reference permission column of Table 2, wherein the reference permission information of the speech text is the content recorded in the reference permission column of Table 2.

[0160] For example, if the speech text is generated for the speech content of Zhang San, since the file level of the speech text of Zhang San is a general file, after obtaining the speech text of Zhang San, only the label corresponding to the general confidential file needs to be set for the speech text of Zhang San, so that all personnel in the company can refer to the speech text of Zhang San.

[0161] For example, if the speech text is generated for the speech content of Li Si, since the file level of the speech text of Li Si is a confidential file, after obtaining the speech text of Li Si, only the label corresponding to the confidential file needs to be set for the speech text of Li Si, so that only the personnel at the level of the department head of the company can refer to the speech text of Li Si.

[0162] In another optional embodiment of the present application, in order to improve the transcription efficiency of the speech, the present application also performs merging processing on the speech texts with high similarity.

[0163] Specifically, the merging process of the speech texts with high similarity includes the following steps S206 to S208:

[0164] Step S206, obtaining the speech text of a second speaker, wherein the speech texts of the second speaker and the first speaker have the same reference permission;

[0165] Step S207, determining the text similarity between the speech text of the first speaker and the speech text of the second speaker;

[0166] Step S208, in the case that the text similarity is greater than a preset similarity threshold, merging the speech text of the first speaker and the speech text of the second speaker to obtain a merged speech text.

[0167] In an optional embodiment of the present application, in order to facilitate the limitation of the text reference permission, only the speech texts of different speakers with the same reference permission are considered for merging.

[0168] Furthermore, before merging the speech texts of different speakers, the similarity of the speech texts of different speakers needs to be considered, and the speech texts with similarity greater than a preset similarity threshold are considered to meet the merging conditions. In an optional embodiment of the present application, the similarity of the speech texts of different speakers is obtained by calculating the feature distance between the different speech texts, for example, by calculating the Euclidean distance between the different texts or calculating the cosine distance between the different texts. This application does not impose any restrictions on this.

[0169] Furthermore, given that the first speaker and the second speaker have the same text review permissions and the text similarity between the speech texts of the first speaker and the second speaker is high, it can be considered that the speech content of the first speaker and the second speaker is related, and the merged speech text can be used as the common speech text of the first speaker and the second speaker.

[0170] For example, let's say the first speaker's speech is in text A.txt and the second speaker's speech is in text B.txt. The access permissions for both speakers are visible to everyone in the company, and if the similarity between speech texts A.txt and B.txt exceeds a preset similarity threshold, then speech texts A.txt and B.txt can be merged to produce speech text C.txt. In this case, speech text C.txt can serve as both the first speaker's and the second speaker's speech texts.

[0171] In an optional implementation of the present application, after obtaining the speech text of the first speaker, the speech text of the first speaker can be sent to the terminal device of the first speaker (for example: the first speaker’s mobile phone, computer, etc.) according to the first speaker’s reference needs, so that the first speaker can reference his speech text.

[0172] Furthermore, in order to facilitate others to review the speech text of the first speaker, after setting the review permission of the speech text of the first speaker, the speech text of the first speaker can be sent to the cloud so that other people who meet the review permission can review the speech text of the first speaker.

[0173] In summary, the voice recognition method provided by the embodiments of the present application determines whether the first speaker corresponding to the speech voice has voice transcription permission through the voiceprint feature in the to-be-recognized speech voice, and then performs voice recognition processing on the speech voice based on the voice transcription permission of the first speaker. Before performing voice recognition on the speech voice of the speaker, the method can automatically determine whether the speaker has voice transcription permission, and only when it is determined that the speaker has voice transcription permission, voice recognition processing is performed on the speech voice of the speaker. The scheme can flexibly determine whether to perform voice recognition processing on the speech voice of the speaker, thereby improving the flexibility of voice recognition and being beneficial to alleviating the problems of voice recognition resource shortage and recognition resource waste.

[0174] Exemplary devices

[0175] Correspondingly, the embodiments of the present application also provide a voice recognition device, please refer to Figure 5 , Figure 5 The structure schematic diagram of the voice recognition device provided by another embodiment of the present application.

[0176] The voice recognition device comprises:

[0177] The extraction unit 501 is configured to extract a voiceprint feature from the to-be-recognized speech voice.

[0178] The judgment unit 502 is configured to determine whether the first speaker corresponding to the speech voice has voice transcription permission according to the voiceprint feature.

[0179] The recognition unit 503 is configured to perform voice recognition processing on the speech voice to obtain the speech text of the first speaker if the first speaker corresponding to the speech voice has voice transcription permission.

[0180] In an optional embodiment of the present application, the judgment unit 501 comprises:

[0181] The information confirmation sub-unit is configured to determine the identity information of the first speaker from the voiceprint library according to the voiceprint feature.

[0182] The permission judgment sub-unit is configured to determine whether the first speaker has voice transcription permission according to a preset voice transcription permission table and the identity information of the first speaker, wherein the preset voice transcription permission table records the voice transcription permissions of a plurality of speakers.

[0183] In an optional embodiment of the present application, the voiceprint library is obtained by the following unit:

[0184] The sample acquisition unit is configured to acquire preset voice sample information of a plurality of speakers.

[0185] a voiceprint extraction unit configured to extract voiceprint features from the speech sample information of the multiple speakers respectively;

[0186] a voiceprint library construction unit configured to construct the voiceprint library based on a correspondence between the voiceprint features in the speech sample information and the identity information of the multiple speakers.

[0187] In an optional implementation of the present application, the speech recognition device further comprises:

[0188] an identity determination unit configured to determine text review identity information of the first speaker, the text review identity information comprising permission information of other people reviewing the speech text of the first speaker;

[0189] a permission setting unit configured to set the review permission of the speech text of the first speaker based on the text review identity information of the first speaker.

[0190] In an optional implementation of the present application, the speech recognition device further comprises:

[0191] a text obtaining unit configured to obtain the speech text of a second speaker, wherein the review permissions of the speech texts of the first speaker and the second speaker are the same;

[0192] a similarity determination unit configured to determine a text similarity between the speech text of the first speaker and the speech text of the second speaker;

[0193] a text merging unit configured to merge the speech text of the first speaker and the speech text of the second speaker to obtain a merged speech text if the text similarity is greater than a preset similarity threshold.

[0194] In an optional implementation of the present application, the recognition unit 503 comprises:

[0195] a retrieval subunit configured to retrieve a personalized transcription model corresponding to the first speaker from a preset personalized transcription resource library according to the voiceprint features of the first speaker; the personalized transcription model is obtained by recognizing and training speech with specific speech characteristics, and the speech of the first speaker has the specific speech characteristics;

[0196] a personalized transcription subunit configured to, if the personalized transcription model corresponding to the first speaker is retrieved from the preset personalized transcription resource library, perform speech recognition processing on the speech of the first speaker by using the personalized transcription model to obtain the speech text of the first speaker.

[0197] In an optional implementation of the present application, the personalized transcription model corresponding to the first speaker is obtained by performing speech recognition training on the speech of the first speaker.

[0198] In an optional embodiment of the present application, the speech recognition device further includes:

[0199] a permission triggering unit, configured to determine whether an adjustment instruction for the voice transcription permission of the first speaker has been obtained;

[0200] If an adjustment instruction for the voice transcription authority of the first speaker is obtained, the voice transcription authority of the first speaker is adjusted according to the adjustment instruction.

[0201] The speech recognition device provided in this embodiment is based on the same concept as the speech recognition method provided in the above embodiments of this application. It can execute the speech recognition method provided in any of the above embodiments of this application and has the corresponding functional modules and beneficial effects of executing the speech recognition method. For technical details not fully described in this embodiment, please refer to the specific processing content of the speech recognition method provided in the above embodiments of this application, and will not be repeated here.

[0202] Exemplary electronic devices

[0203] Another embodiment of the present application further provides an electronic device, see Figure 6 , Figure 6 This is a schematic diagram of the electronic device structure provided by another embodiment of the present application, such as Figure 6 As shown, the device includes:

[0204] Memory 200 and processor 210;

[0205] The memory 200 is connected to the processor 210 and is used to store programs;

[0206] The processor 210 is configured to implement the speech recognition method disclosed in any of the above embodiments by running the program stored in the memory 200 .

[0207] Specifically, the electronic device may further include: a bus, a communication interface 220 , an input device 230 and an output device 240 .

[0208] The processor 210, the memory 200, the communication interface 220, the input device 230 and the output device 240 are interconnected via a bus.

[0209] A bus may include a pathway that transfers information between components of a computer system.

[0210] Processor 210 can be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, or the like, or an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present invention. Alternatively, it can be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware components.

[0211] The processor 210 may include a main processor, and may also include a baseband chip, a modem, and the like.

[0212] The memory 200 stores a program for executing the technical solution of the present invention, and may also store an operating system and other key services. Specifically, the program may include program code, which includes computer operating instructions. More specifically, the memory 200 may include read-only memory (ROM), other types of static storage devices that can store static information and instructions, random access memory (RAM), other types of dynamic storage devices that can store information and instructions, disk storage, flash memory, etc.

[0213] The input device 230 may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.

[0214] Output device 240 may include devices that allow information to be output to a user, such as a display screen, printer, speakers, etc.

[0215] The communication interface 220 may include any device such as a transceiver to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0216] The processor 210 executes the program stored in the memory 200 and calls other devices, which can be used to implement the various steps of any speech recognition method provided in the above embodiments of the present application.

[0217] Exemplary computer program products and storage media

[0218] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the speech recognition method according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0219] The computer program product may be written in any combination of one or more programming languages ​​to implement the program code for performing the operations of the embodiments of the present application, including object-oriented programming languages ​​such as Java, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0220] In addition, an embodiment of the present application may also be a storage medium having a computer program stored thereon, and the computer program is used by a processor to execute the steps of the speech recognition method according to various embodiments of the present application described in the "Exemplary Method" section above of this specification, specifically the following steps:

[0221] Step S201, extracting voiceprint features from the speech to be recognized;

[0222] Step S202: determining, based on the voiceprint features, whether the first speaker corresponding to the speech has voice transcription authority;

[0223] Step S203: If the first speaker corresponding to the speech voice has the speech transcription permission, speech recognition processing is performed on the speech voice to obtain the speech text of the first speaker.

[0224] For the sake of simplicity, the aforementioned method embodiments are described as a series of action combinations. However, those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0225] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similarities between the various embodiments can be referred to in conjunction with each other. For device embodiments, since they are generally similar to method embodiments, their description is relatively simple, and for relevant details, reference can be made to the description of the method embodiments.

[0226] The steps in the methods of each embodiment of the present application can be adjusted in sequence, merged, and deleted according to actual needs, and the technical features recorded in each embodiment can be replaced or combined.

[0227] The modules and sub-modules in the devices and terminals of the various embodiments of the present application can be merged, divided, and deleted according to actual needs.

[0228] In the several embodiments provided in this application, it should be understood that the disclosed terminals, devices, and methods can be implemented in other ways. For example, the terminal embodiments described above are merely illustrative. For example, the division of modules or submodules is merely a logical function division. In actual implementation, there may be other division methods, such as multiple submodules or modules can be combined or integrated into another module, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or module, which can be electrical, mechanical or other forms.

[0229] The modules or submodules described as separate components may or may not be physically separate, and the components of the modules or submodules may or may not be physical modules or submodules, that is, they may be located in one place or distributed across multiple network modules or submodules. Some or all of the modules or submodules may be selected to achieve the purpose of this embodiment according to actual needs.

[0230] In addition, each functional module or submodule in each embodiment of the present application may be integrated into a processing module, or each module or submodule may exist physically separately, or two or more modules or submodules may be integrated into a single module. The above-mentioned integrated modules or submodules may be implemented in the form of hardware or software functional modules or submodules.

[0231] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0232] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, software units executed by a processor, or a combination of the two. The software units may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0233] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0234] The above description of the disclosed embodiments will enable those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is to be construed in the widest manner consistent with the principles and novel features disclosed herein.

Claims

1. A speech recognition method, characterized in that: include: Extracting voiceprint features from the speech to be recognized; Determining, based on the voiceprint features, whether the first speaker corresponding to the speech has voice transcription authority; If the first speaker corresponding to the speech voice has speech transcription permission, performing speech recognition processing on the speech voice to obtain the speech text of the first speaker; Obtaining a speech text of the second speaker, wherein the first speaker and the second speaker have the same access rights to the speech texts; Determining text similarity between a speech text of the first speaker and a speech text of the second speaker; When the text similarity is greater than a preset similarity threshold, the speech text of the first speaker and the speech text of the second speaker are merged to obtain a merged speech text.

2. The method according to claim 1, characterized in that The determining, based on the voiceprint feature, whether the first speaker corresponding to the speech has voice transcription authority includes: determining the identity information of the first speaker from a voiceprint database according to the voiceprint feature; According to a preset voice transcription permission table and the identity information of the first speaker, it is determined whether the first speaker has the voice transcription permission; wherein the preset voice transcription permission table records the voice transcription permissions of multiple speakers.

3. The method according to claim 2, characterized in that The voiceprint database is obtained in the following way: Obtaining preset multiple speaker voice sample information; respectively extracting voiceprint features from the speech sample information of the plurality of speakers; The voiceprint database is constructed based on the correspondence between the voiceprint features in the voice sample information and the identity information of the multiple speakers.

4. The method according to claim 1, wherein Also includes: Determining text access identification information of the first speaker, where the text access identification information includes permission information for other persons to access the speech text of the first speaker; Based on the text review identification information of the first speaker, review authority for the speech text of the first speaker is set.

5. The method according to claim 1, wherein The performing speech recognition processing on the speech to obtain the speech text of the first speaker includes: Retrieving a personalized transcription model corresponding to the first speaker from a preset personalized transcription resource library based on the voiceprint features of the first speaker; the personalized transcription model is obtained by performing recognition training on speech with specific speech characteristics, and the speech of the first speaker has the specific speech characteristics; If a personalized transcription model corresponding to the first speaker is retrieved from a preset personalized transcription resource library, speech recognition processing is performed on the speech of the first speaker using the personalized transcription model to obtain a speech text of the first speaker.

6. The method according to claim 5, characterized in that The personalized transcription model corresponding to the first speaker is obtained by performing speech recognition training on the speech of the first speaker.

7. The method according to claim 1, characterized in that Also includes: Determining whether an adjustment instruction for the voice transcription authority of the first speaker is obtained; If an adjustment instruction for the voice transcription authority of the first speaker is obtained, the voice transcription authority of the first speaker is adjusted according to the adjustment instruction.

8. A speech recognition device, characterized in that: include: An extraction unit, configured to extract voiceprint features from the speech to be recognized; a judgment unit, configured to judge, based on the voiceprint feature, whether the first speaker corresponding to the speech has the voice transcription permission; a recognition unit configured to perform speech recognition processing on the speech to obtain a speech text of the first speaker if the first speaker corresponding to the speech has speech transcription authority; Obtaining a speech text of the second speaker, wherein the first speaker and the second speaker have the same access rights to the speech texts; Determining text similarity between a speech text of the first speaker and a speech text of the second speaker; When the text similarity is greater than a preset similarity threshold, the speech text of the first speaker and the speech text of the second speaker are merged to obtain a merged speech text.

9. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the speech recognition method according to any one of claims 1 to 8 by running instructions in the memory.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the speech recognition method according to any one of claims 1 to 8 is executed.

Images