Speech recognition method and device, electronic equipment, medium and computer program product

Abstract

The invention provides a voice recognition method and device, electronic equipment, a medium and a computer program product, and relates to the technical field of voice recognition, and the method comprises the steps: collecting a target voice signal of a target user, processing the target voice signal at the same time through a plurality of decoding networks, obtaining a plurality of recognition results, and according to the plurality of recognition results, carrying out the processing of the target voice signal through a plurality of decoding networks. The target recognition result corresponding to the target voice signal is obtained, the target voice signal is processed at the same time through multiple decoding networks, the target recognition result can be quickly obtained subsequently, the speed of voice recognition is improved, and when the method is applied to an intelligent cabin scene of a vehicle, waiting time of passengers in a cabin is shortened through quick voice recognition, and the efficiency of voice recognition is improved. Therefore, the user experience can be improved, the final recognition result is obtained through the multiple recognition results, the accuracy of voice recognition is improved, and the user experience can be further improved.

Description

Technical Field

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

[0002] With the development of technology, more and more smart devices are entering our lives. Most of these devices can be controlled by voice, greatly facilitating our daily lives. During voice control, speech recognition technology is applied. Currently, speech recognition is typically performed using speech recognition-related models or networks. To improve recognition accuracy, the current method is to process the user's speech sequentially through multiple models or networks. However, this method is slow. Summary of the Invention

[0003] To overcome the problems existing in the related art, the present disclosure provides a speech recognition method, apparatus, electronic device, medium and computer program product.

[0004] According to a first aspect of an embodiment of the present disclosure, a speech recognition method is provided, comprising: collecting a target speech signal of a target user; simultaneously processing the target speech signal through multiple decoding networks to obtain multiple recognition results; and obtaining a target recognition result corresponding to the target speech signal based on the multiple recognition results.

[0005] Optionally, the target speech signal is processed simultaneously by multiple decoding networks to obtain multiple recognition results, including: encoding the target speech signal through an encoder to obtain an encoded target speech signal; and processing the encoded target speech signal through the multiple decoding networks to obtain the multiple recognition results.

[0006] Optionally, the encoded target speech signal includes a filter bank FBANK feature, and encoding the target speech signal through an encoder to obtain the encoded target speech signal includes: processing the target speech signal to obtain Mel-frequency cepstral coefficient MFCC acoustic features; inputting the MFCC acoustic features into the encoder for encoding to obtain FBANK features.

[0007] Optionally, the multiple decoding networks include a first decoding network and a second decoding network, the first decoding network is obtained by training the voice signals of multiple users, and the second decoding network is obtained by training the voice signals of the target user; the simultaneously processing the target voice signal through multiple decoding networks to obtain multiple recognition results includes: processing the target voice signal through the first decoding network to obtain a first recognition result, and simultaneously processing the target voice signal through the second decoding network to obtain a second recognition result; obtaining the target recognition result corresponding to the target voice signal based on the multiple recognition results includes: obtaining the target recognition result corresponding to the target voice signal based on the first recognition result and the second recognition result.

[0008] Optionally, the first decoding network includes a first sub-decoding network for processing characters corresponding to the speech signal and a second sub-decoding network for processing syllables corresponding to the speech signal; the processing of the target speech signal by the first decoding network to obtain a first recognition result includes: processing the target speech signal by the first sub-decoding network to obtain a first sub-recognition result, and at the same time processing the target speech signal by the second sub-decoding network to obtain a second sub-recognition result, and using the first sub-recognition result and the second sub-recognition result as the first recognition result.

[0009] Optionally, the second decoding network includes a third sub-decoding network for processing syllables corresponding to the speech signal; processing the target speech signal through the second decoding network to obtain a second recognition result includes: processing the target speech signal through the third sub-decoding network to obtain the second recognition result.

[0010] Optionally, obtaining the target recognition result corresponding to the target speech signal based on the first recognition result and the second recognition result includes: obtaining a first acoustic score for evaluating the second sub-recognition result, and obtaining a second acoustic score for evaluating the second recognition result; when the second acoustic score is higher than or equal to the first acoustic score, using the second recognition result as the target recognition result corresponding to the target speech signal; when the second acoustic score is lower than the first acoustic score, using the first sub-recognition result as the target recognition result corresponding to the target speech signal.

[0011] Optionally, the method further includes: after updating the first decoding network, and when it is determined that a subnetwork in a second decoding network located in the first server and the first decoding network is different from a subnetwork in the second server, compiling the second decoding network by the second server to obtain a compiled second decoding network; and pushing the compiled second decoding network to the first server by the second server, and overwriting the second decoding network in the first server.

[0012] Optionally, the second decoding network is compiled in the following manner: obtaining a voice sample corresponding to a target user; processing the voice sample into a label sequence; inputting the label sequence into a personalized decoding network to compile a subnetwork in the personalized decoding network to obtain a compiled second decoding network.

[0013] According to a second aspect of an embodiment of the present disclosure, a speech recognition device is provided, comprising: an acquisition module for acquiring a target speech signal of a target user; a recognition module for simultaneously processing the target speech signal through multiple decoding networks to obtain multiple recognition results; and an acquisition module for obtaining a target recognition result corresponding to the target speech signal based on the multiple recognition results.

[0014] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to implement the steps of the method described in the first aspect when executing the instructions.

[0015] According to a fourth aspect of an embodiment of the present disclosure, a vehicle is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to implement the steps of the method described in the first aspect when executing the instructions.

[0016] According to a fifth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the program instructions are executed by a processor, the steps of the method described in the first aspect of the present disclosure are implemented.

[0017] According to a sixth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising a computer program, which implements the steps of the method described in the first aspect when executed by a processor.

[0018] The speech recognition method, device, electronic device, medium and computer program product provided by the present disclosure first collect the target speech signal of the target user, then process the target speech signal simultaneously through multiple decoding networks to obtain multiple recognition results, and then obtain the target recognition result corresponding to the target speech signal based on the multiple recognition results. The target speech signal is processed simultaneously by multiple decoding networks, which facilitates the subsequent rapid acquisition of the target recognition result and improves the speech recognition rate. When applied to the intelligent cockpit scenario of a vehicle, the waiting time of the passengers in the cockpit is reduced through rapid speech recognition, which can improve the user experience. In addition, the final recognition result is obtained through multiple recognition results to improve the accuracy of speech recognition, which can further improve the user experience.

[0019] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0021] Figure 1 It is a schematic diagram of a speech recognition method in the prior art;

[0022] Figure 2 is a flow chart of a speech recognition method according to an exemplary embodiment of the present disclosure;

[0023] Figure 3 is a flow chart of a speech recognition method according to an exemplary embodiment of the present disclosure;

[0024] Figure 4 is a flow chart of a speech recognition method according to an exemplary embodiment of the present disclosure;

[0025] Figure 5 is a flow chart of a speech recognition method according to an exemplary embodiment of the present disclosure;

[0026] Figure 6 is a block diagram of a speech recognition device according to an exemplary embodiment;

[0027] Figure 7 The figure is a block diagram of an electronic device for a voice recognition method according to an exemplary embodiment. DETAILED DESCRIPTION

[0028] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0029] The embodiments described in the following examples of the present disclosure do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0030] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the corresponding data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0031] With the development of science and technology, more and more smart devices have entered people's lives. Most smart devices have realized voice control, which has greatly facilitated people's lives. For example, smart devices can be navigation devices in vehicles. In the process of voice control, speech recognition technology is applied. Currently, speech recognition-related models or networks are usually used for speech recognition. In order to improve recognition accuracy, the current method is to process the user's speech in sequence through multiple models or networks. For example, please refer to Figure 1 The multiple networks may include a first decoding network and a second decoding network. The speech signal to be recognized is input into the first decoding network, which then recognizes the speech signal and obtains an output result. This output result is then input into the second decoding network to further recognize the speech signal and obtain the final decoding result. However, the inventors have found that this method is slow in speech recognition. Furthermore, the existing decoding method uses a non-end-to-end model, which has low recognition accuracy.

[0032] To solve the above problems, the present disclosure provides a speech recognition method, see Figure 2 , the speech recognition method can be applied to Figure 6 The speech recognition device 200 shown, Figure 7 The electronic device 800, computer readable storage medium and computer program product shown in the embodiment are used as an example. The electronic device can be a navigation device, audio device, lighting device and other vehicle-mounted devices in the smart cockpit of the vehicle. The electronic device can also be a voice assistant, smart home device, etc. Figure 2 The process shown in FIG. 1 is described in detail. Specifically, the speech recognition method may include the following steps:

[0033] Step S110: collecting a target voice signal of a target user.

[0034] The target voice signal of the target user is collected by a sound pickup device on the electronic device or a sound pickup device matched with the electronic device. The target voice signal is the sound signal that needs to be recognized.

[0035] Step S120: Process the speech signal simultaneously through multiple decoding networks to obtain multiple recognition results.

[0036] Decoding networks are used to convert sound signals into text sequences, thereby achieving speech recognition. In this disclosure, multiple decoding networks simultaneously process speech signals in parallel, improving the processing speed of speech signals. Each of the multiple decoding networks outputs a recognition result, resulting in multiple recognition results. Each recognition result refers to the text converted by the corresponding decoding network after the target speech signal is converted.

[0037] For example, see Figure 3 , taking the example of multiple decoding networks including a first decoding network and a second decoding network, the multiple decoding networks include a first decoding network and a second decoding network. Among them, the first decoding network is obtained through training of voice signals of multiple users, and is a decoding network common to all users. The second decoding network is obtained through training of the voice signal of the target user, and is a decoding network trained for the target user. Voice recognition is performed through the second decoding network, and the recognition result obtained is more in line with the user habits of the target user. Optionally, the target account logged in by the target user through the electronic device is obtained, and the second decoding network corresponding to the target account is obtained. In one embodiment, the target voice signal is processed by the first decoding network to obtain a first recognition result, and the target voice signal is processed by the second decoding network to obtain a second recognition result. The first decoding network and the second decoding network recognize the target voice signal at the same time, which can improve the speed of voice recognition.

[0038] As one approach, the first decoding network includes a first sub-decoding network for processing characters corresponding to the speech signal and a second sub-decoding network for processing syllables corresponding to the speech signal. It is understood that the first sub-decoding network is a decoding network that takes characters of the speech signal as basic input, and the second sub-decoding network is a decoding network that takes syllables of the speech signal as basic input. A first recognition result is obtained by processing the target speech signal through the first sub-decoding network to obtain a first sub-recognition result, while simultaneously processing the target speech signal through the second sub-decoding network to obtain a second sub-recognition result, and using the first sub-recognition result and the second sub-recognition result as the first recognition result.

[0039] In one embodiment, the second decoding network includes a third sub-decoding network for processing syllables corresponding to the speech signal. It is understood that the third sub-decoding network is a decoding network that uses syllables of the speech signal as basic input. The second recognition result is obtained by processing the target speech signal through the third sub-decoding network to obtain the second recognition result.

[0040] It should be noted that when the first sub-decoding network, the second sub-decoding network and the third sub-decoding network exist at the same time, the three decoding networks process the speech signal at the same time to obtain their respective corresponding recognition results.

[0041] In another embodiment, the target speech signal is encoded by an encoder to obtain an encoded target speech signal. Exemplarily, the encoded target speech signal includes FBANK (FilterBank) features. The target speech signal is processed to obtain MFCC (Mel-Frequency Cepstral Coefficients) acoustic features; the MFCC acoustic features are input into the encoder for encoding to obtain FBANK features. The encoded target speech signal (i.e., FBANK features) is then processed by the multiple decoding networks to obtain the multiple recognition results. Continuing with the example that the first decoding network includes a first sub-decoding network and a second sub-decoding network, and the second decoding network includes a third sub-decoding network, the FBANK features are simultaneously processed by the first sub-decoding network, the second sub-decoding network, and the third sub-decoding network to obtain a first sub-recognition result output by the first sub-decoding network, a second sub-recognition result output by the second sub-decoding network, and a second recognition result output by the third sub-decoding network.

[0042] Step S130: Obtain a target recognition result corresponding to the target speech signal based on the multiple recognition results.

[0043] In this step, multiple recognition results are integrated to obtain a target recognition result corresponding to the target speech signal. The final target recognition result has a high accuracy rate, thereby improving the effect of speech recognition.

[0044] In one embodiment, a target recognition result corresponding to the target speech signal is obtained based on the first recognition result and the second recognition result.

[0045] As a method, taking the first recognition result including the first sub-recognition result and the second sub-recognition result as an example, a first acoustic score for evaluating the second sub-recognition result is obtained, and a second acoustic score for evaluating the second recognition result is obtained. When the second acoustic score is higher than or equal to the first acoustic score, the second recognition result is used as the target recognition result corresponding to the target speech signal. Otherwise, when the second acoustic score is lower than the first acoustic score, the first sub-recognition result is used as the target recognition result corresponding to the target speech signal.

[0046] As another example, taking the case where the first recognition result includes a first sub-recognition result and a second sub-recognition result, a first similarity between the first sub-recognition result and the second recognition result is obtained, a second similarity between the second sub-recognition result and the second recognition result is obtained, and a third similarity between the first sub-recognition result and the second sub-recognition result is obtained. The largest similarity among the first, second, and third similarities is selected as the target similarity, and any one of the two recognition results corresponding to the target similarity is used as the target recognition result. For example, if the third similarity is the largest among the three similarities, then the third similarity is used as the target similarity, and any one of the first sub-recognition result and the second sub-recognition result corresponding to the third similarity is used as the target recognition result.

[0047] The speech recognition method provided in this embodiment first collects the target speech signal of the target user, then processes the target speech signal simultaneously through multiple decoding networks to obtain multiple recognition results, and then obtains the target recognition result corresponding to the target speech signal based on the multiple recognition results. The target speech signal is processed simultaneously by multiple decoding networks, which facilitates the subsequent rapid acquisition of the target recognition result and improves the speech recognition rate. When applied to the intelligent cockpit scenario of a vehicle, the waiting time of the passengers in the cockpit can be reduced through rapid speech recognition, which can improve the user experience. In addition, the final recognition result is obtained through multiple recognition results to improve the accuracy of speech recognition, which can further improve the user experience.

[0048] The first decoding network and the second decoding network are both located in the first server. In order to improve the accuracy of speech recognition, the encoding network involved in the present disclosure needs to be updated. Optionally, the first decoding network can be updated regularly. After the first decoding network is updated, and when it is determined that the subnetwork in the second decoding network located in the first server and the subnetwork in the second server are different, the second decoding network is compiled by the second server to obtain a compiled second decoding network, wherein the second decoding network can be composed of three subnetworks: T.fst (referring to the acoustic model), L.fst (referring to the sound and language model), and G.fst (referring to the language model). Exemplarily, at both ends of the first server and the second server, when at least one subnetwork among T.fst, L.fst and G.fst is different, the second decoding network needs to be compiled. For example, when T.fst in the first server is different from T.fst in the second server, and L.fst in the first server is different from L.fst in the second server, the second decoding network needs to be compiled. The compiled second decoding network is pushed to the first server through the second server, and covers the second decoding network in the first server.

[0049] In this embodiment, speech recognition is performed on the target speech signal in the first server through the first decoding network and the second decoding network, while the compilation of the second decoding network is performed in the second server, so that the compilation process and the recognition process in the two servers will not affect each other, thereby improving the stability of the overall architecture.

[0050] In this disclosure, compilation refers to the process of synthesizing T.fst, L.fst, and G.fst to obtain an updated second decoding network. Optionally, the second decoding network is compiled in the following manner: first, a speech sample corresponding to the target user is obtained. As one approach, the target user's navigation history is obtained, where the navigation history may include favorite addresses and historical navigation addresses. A sentence pattern stored in the database of the first server is obtained through a second server, and a sentence is constructed based on the navigation history. For example, if the historical navigation address is Central Park, the sentence constructed based on the stored sentence pattern may be "I want to navigate to Central Park now." A navigation speech signal is thereby obtained, and this navigation speech signal is used as the speech sample. As another approach, the target user's historical input speech signals are stored and used as the speech sample. The speech sample is then processed into a label sequence, for example, a CTC (Connectionist Temporal Classification) label sequence. Finally, the label sequence is input into the personalized decoding network to compile the subnetworks within the personalized decoding network, thereby obtaining a compiled second decoding network. Exemplarily, the label sequence is input into T.fst to obtain a modeling unit; the modeling unit is then input into L.fst to obtain an output word; the output word is input into G.fst to obtain a target word; based on T.fst, L.fst and G.fst, a second decoding network is synthesized to achieve compilation of the second decoding network.

[0051] Optionally, the encoding networks provided in the present disclosure are all end-to-end models. Compared with the non-end-to-end models in the related art, the end-to-end models of the present disclosure can make the speech recognition process more automated and efficient.

[0052] Optionally, the encoding network in the present disclosure may be an ngram language model.

[0053] It should be noted that the second coding network disclosed in the present invention is targeted at the target users of the electronic device and has stronger personalized features. Therefore, by adjusting the second coding network, the coding network can be expanded to richer scenarios. For example, in the audio scene of the smart cockpit, the second coding network is trained by the corpus associated with the target user and the audio, so the coding network can be applicable to the audio scene. For another example, in the call scene of the smart cockpit, the second coding network is trained by the corpus associated with the target user and the call, so the coding network can be applicable to the call scene. For another example, in the home scene, the second coding network is trained by the corpus associated with the target user and the home, so the coding network can be applicable to the home scene.

[0054] The present disclosure provides a speech recognition method, which is applied to the navigation scene of the intelligent cockpit of a vehicle. Figure 5 , the method includes: obtaining a user request, wherein the user request includes a target account logged in by the target user through the vehicle, and a target voice signal of the target user collected in the smart cockpit. Obtaining a second decoding network corresponding to the target account from the cloud database, and obtaining a universal first sub-decoding network and a second sub-decoding network. Encoding the target voice signal in the user request through an encoder to obtain an encoded target voice signal, inputting the encoded target voice signal into the first sub-decoding network, the second sub-decoding network, and the third sub-decoding network respectively for simultaneous processing, obtaining a first sub-recognition result output by the first sub-decoding network, a second sub-recognition result output by the second sub-decoding network, and a second recognition result output by the third sub-decoding network. Based on the first sub-recognition result, the second sub-recognition result, and the second recognition result, a final target recognition result is obtained, thereby completing voice recognition for the target voice signal.

[0055] To improve recognition performance, the decoding network needs to be updated. When the common first and second sub-decoding networks are updated, the L.fst and T.fst in the first and second servers are determined to be the same. Specifically, the L.fst in the first and second servers are determined to be the same, and the T.fst in the first and second servers are determined to be the same. If either L.fst or T.fst is different, the second decoding network is compiled using an offline compilation service. Specifically, the second decoding network is compiled using the target user's navigation history by the second server to obtain a compiled second decoding network. If both L.fst and T.fst are the same, the second decoding network does not need to be compiled.

[0056] The present disclosure provides a speech recognition method that adapts to an end-to-end model rather than a traditional non-end-to-end model, and uses different levels of decoding networks for parallel recognition, thereby realizing the user's personalized navigation recognition needs, and also making the recognition results more accurate and the recognition efficiency higher. By obtaining the navigation address and favorite address of the target user every day, the user navigation corpus can be updated, and the second decoding network can be updated through the second server, so that the update of the second decoding network with personalized navigation recognition is more timely. The offline compilation service and the recognition service do not affect each other, that is, the offline compilation service is in the second server, and the recognition service is in the first server. The separation of the two parts makes the overall architecture more stable.

[0057] The present disclosure also provides a vehicle, a vehicle processor; and a memory for storing processor-executable instructions; wherein the processor is configured to implement the following steps when executing the instructions: collecting a target voice signal of a target user in a vehicle cabin; simultaneously processing the target voice signal through multiple decoding networks to obtain multiple recognition results; and obtaining a target recognition result corresponding to the target voice signal based on the multiple recognition results.

[0058] Exemplarily, the multiple decoding networks include a first decoding network and a second decoding network. The first decoding network is trained using speech signals from multiple users. It is understood that the first decoding network is a universal network, while the second decoding network is trained using speech signals from a target user, specifically trained for the target user. The first and second decoding networks simultaneously recognize the target speech signal, obtaining a first recognition result corresponding to the first decoding network and a second recognition result corresponding to the second decoding network.

[0059] The second decoding network is trained based on the user habits of the target user, and the target account of the target user logged in through the vehicle can be obtained, and the second decoding network is obtained based on the target account.

[0060] Based on the same inventive concept, the present disclosure provides a speech recognition device 200, see Figure 6 , the speech recognition device 200 includes:

[0061] The acquisition module 210 is used to acquire the target voice signal of the target user;

[0062] The recognition module 220 is configured to simultaneously process the target speech signal through multiple decoding networks to obtain multiple recognition results;

[0063] The acquisition module 230 is configured to obtain a target recognition result corresponding to the target speech signal based on the multiple recognition results.

[0064] Optionally, the identification module 220 includes:

[0065] An encoding module, configured to encode the target speech signal through an encoder to obtain an encoded target speech signal;

[0066] The first recognition module is configured to process the encoded target speech signal through the multiple decoding networks to obtain the multiple recognition results.

[0067] Optionally, the encoded target speech signal includes a filter bank FBANK feature, and the encoding module includes:

[0068] The first encoding module is used to process the target speech signal to obtain Mel-frequency cepstral coefficient (MFCC) acoustic features;

[0069] The second encoding module is used to input the MFCC acoustic features into the encoder for encoding to obtain FBANK features.

[0070] Optionally, the multiple decoding networks include a first decoding network and a second decoding network, the first decoding network is obtained by training with speech signals of multiple users, and the second decoding network is obtained by training with a speech signal of the target user; the recognition module 220 includes:

[0071] The second recognition module is used to process the target speech signal through the first decoding network to obtain a first recognition result, and at the same time process the target speech signal through the second decoding network to obtain a second recognition result.

[0072] The acquisition module 230 includes:

[0073] The target result acquisition module is used to obtain a target recognition result corresponding to the target speech signal according to the first recognition result and the second recognition result.

[0074] Optionally, the first decoding network includes a first sub-decoding network for processing characters corresponding to the speech signal and a second sub-decoding network for processing syllables corresponding to the speech signal, and a second recognition module, which is specifically used to process the target speech signal through the first sub-decoding network to obtain a first sub-recognition result, and at the same time process the target speech signal through the second sub-decoding network to obtain a second sub-recognition result, and use the first sub-recognition result and the second sub-recognition result as the first recognition result.

[0075] Optionally, the second decoding network includes a third sub-decoding network for processing syllables corresponding to the speech signal, and a second recognition module is specifically used to process the target speech signal through the third sub-decoding network to obtain the second recognition result.

[0076] Optionally, the target result acquisition module is specifically used to obtain a first acoustic score for evaluating the second sub-recognition result, and to obtain a second acoustic score for evaluating the second recognition result; when the second acoustic score is higher than or equal to the first acoustic score, the second recognition result is used as the target recognition result corresponding to the target speech signal; when the second acoustic score is lower than the first acoustic score, the first sub-recognition result is used as the target recognition result corresponding to the target speech signal.

[0077] Optionally, the speech recognition device 200 further includes:

[0078] a compiling module configured to, after updating the first decoding network and upon determining that a subnetwork in a second decoding network located in the first server and the subnetwork in the second server are different, compile the second decoding network through the second server to obtain a compiled second decoding network;

[0079] A pushing module is configured to push the compiled second decoding network to the first server through the second server, and cover the second decoding network in the first server.

[0080] Optionally, the compilation module includes:

[0081] The sample acquisition module is used to obtain the voice sample corresponding to the target user;

[0082] A serialization module, configured to process the speech sample into a label sequence;

[0083] The compiling submodule is configured to input the label sequence into a personalized decoding network to compile a subnetwork in the personalized decoding network to obtain a compiled second decoding network.

[0084] Regarding the speech recognition device 200 in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0085] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, which implement the steps of the speech recognition method provided by the present disclosure when the program instructions are executed by a processor.

[0086] Figure 7 8 is a block diagram of an electronic device for a speech recognition method according to an exemplary embodiment. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, a vehicle, a smart home device, etc.

[0087] Reference Figure 7 , the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output interface 812 , a sensor component 814 , and a communication component 816 .

[0088] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 802 may include one or more modules to facilitate interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.

[0089] The memory 804 is configured to store various types of data to support operations on the electronic device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0090] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0091] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor can not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have a focal length and optical zoom capability.

[0092] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), which is configured to receive external audio signals when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0093] The input / output interface 812 provides an interface between the processing component 802 and peripheral interface modules, such as a keyboard, a click wheel, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0094] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the electronic device 800, the relative positioning of components, such as the display and keypad of the electronic device 800. The sensor assembly 814 can also detect changes in the position of the electronic device 800 or a component of the electronic device 800, the presence or absence of user contact with the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and temperature changes of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0095] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0096] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.

[0097] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by the processor 820 of the electronic device 800 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0098] In another exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program that can be executed by a programmable device, and the computer program has a code portion for performing the above-mentioned speech recognition method when executed by the programmable device.

[0099] Those skilled in the art will also appreciate that the various illustrative logical blocks and steps listed in the embodiments of the present application can be implemented by electronic hardware, computer software, or a combination of both. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of the present application.

[0100] In the foregoing detailed description, reference is made to the accompanying drawings, which illustrate, by way of illustration, specific aspects of the present disclosure in which it may be practiced. In this regard, terms indicating directions or expressing positional relationships, such as "center," "longitudinal," "lateral," "length," "width," "thickness," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," "clockwise," "counterclockwise," "axial," "radial," "circumferential," etc., may be used with reference to the orientation of the figures being described. Since the components of the described devices may be positioned in a plurality of different orientations, the directional terms may be used for illustrative purposes rather than restrictive. It should be understood that other aspects may be utilized and structural or logical changes may be made without departing from the concepts of the present disclosure. Therefore, the following detailed description should not be taken in a limiting sense.

[0101] It should be understood that, unless otherwise specifically noted, the features of the various embodiments of the present disclosure described herein may be combined with each other. As used herein, the term "and / or" includes any one of the relevant listed items and any combination of any two or more thereof; similarly, "at least one of" includes any one of the relevant listed items and any combination of any two or more thereof.

[0102] It should be understood that, unless otherwise expressly specified or limited, the terms "join," "attach," "install," "connect," "connect," "fix," etc. used in the embodiments of the present disclosure should be understood in a broad sense. For example, they can be fixedly connected, detachably connected, or integrated; they can be mechanically connected, electrically connected, or communicable with each other; they can be directly connected, or indirectly connected through an intermediate medium, and they can be internally connected between two elements or an interactive relationship between two elements, unless otherwise expressly limited. For those skilled in the art, the specific meanings of the above terms in this article can be understood according to specific circumstances.

[0103] In addition, the term "over" as used in reference to a component, element, or material layer being formed "over" or located "over" a surface may be used herein to mean that the component, element, or material layer is "indirectly" positioned (e.g., placed, formed, deposited, etc.) on the surface such that one or more additional components, elements, or layers are disposed between the surface and the component, element, or material layer. However, the term "over" as used in reference to a component, element, or material layer being formed "over" or located "over" a surface may alternatively have a specific meaning: the component, element, or material layer is "directly" positioned (e.g., placed, formed, deposited, etc.) on the surface, e.g., in direct contact with the surface.

[0104] Although terms such as "first", "second" and "third" may be used herein to describe various components, parts, regions, layers or sections, these components, parts, regions, layers or sections are not limited to these terms. On the contrary, these terms are only used to distinguish one component, part, region, layer or section from another component, part, region, layer or section. Therefore, without departing from the teachings of each example, the first component, part, region, layer or section mentioned in the examples described herein may also be referred to as the second component, part, region, layer or section. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, the features defined as "first" and "second" can explicitly or implicitly include at least one such feature. In the description herein, the meaning of "multiple" is at least two, for example, two, three, etc., unless otherwise clearly and specifically defined.

[0105] It should be understood that spatially relative terms such as "above," "upper," "below," and "lower" are used herein to describe the relationship of one element to another element shown in the figures. In addition to the orientation depicted in the figures, such spatially relative terms are intended to encompass different orientations of the device in use or operation. For example, if the device in the figures is flipped, an element described as being "above" or "upper" relative to another element will be "below" or "lower" relative to the other element. Thus, the term "above" encompasses both above and below orientations, depending on the spatial orientation of the device. The device may be oriented in other ways (e.g., rotated 90 degrees or in other orientations), and the spatially relative terms used herein should be interpreted accordingly.

[0106] Furthermore, the word "exemplary" is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as "exemplary" is not necessarily to be construed as advantageous over other aspects or designs. Rather, the use of the word exemplary is intended to present concepts in a concrete manner. As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified or clear from the context, "X applies to A or B" is intended to mean any of the natural inclusive permutations. That is, if X applies to A; X applies to B; or X applies to both A and B, then "X applies to A or B" satisfies any of the aforementioned instances. Furthermore, the articles "a" and "an," as used in this application and the appended claims, are generally understood to mean "one or more," unless otherwise specified or clear from the context to refer to the singular form.

[0107] Likewise, although the present disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art after reading and understanding the specification and drawings. The present disclosure includes all such modifications and variations and is limited only by the scope of the claims. In particular, with respect to the various functions performed by the components described above (e.g., elements, resources, etc.), unless otherwise indicated, the terms used to describe such components are intended to correspond to any component (functionally equivalent) that performs the specific functions of the described components, even if structurally not equivalent to the disclosed structures. In addition, although specific features of the present disclosure may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations as may be desired and beneficial for any given or specific application. In addition, with respect to the terms "including," "having," "having," "having," or variations thereof used in the specific embodiments or claims, such terms are intended to be inclusive in a manner similar to the term "comprising."

[0108] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the appended claims.

[0109] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A speech recognition method, characterized in that: The method comprises: Collecting target voice signals of target users; The target speech signal is processed simultaneously by multiple decoding networks to obtain multiple recognition results; A target recognition result corresponding to the target speech signal is obtained based on the multiple recognition results.

2. The method according to claim 1, characterized in that The target speech signal is processed simultaneously by multiple decoding networks to obtain multiple recognition results, including: Encoding the target speech signal through an encoder to obtain an encoded target speech signal; The encoded target speech signal is processed by the multiple decoding networks to obtain the multiple recognition results.

3. The method according to claim 2, characterized in that The encoded target speech signal includes a filter bank FBANK feature, and encoding the target speech signal through an encoder to obtain the encoded target speech signal includes: Processing the target speech signal to obtain Mel-frequency cepstral coefficients (MFCC) acoustic features; The MFCC acoustic features are input into the encoder for encoding to obtain FBANK features.

4. The method according to any one of claims 1 to 3, characterized in that The multiple decoding networks include a first decoding network and a second decoding network, the first decoding network is obtained by training with speech signals of multiple users, and the second decoding network is obtained by training with a speech signal of the target user; The target speech signal is processed simultaneously by multiple decoding networks to obtain multiple recognition results, including: Processing the target speech signal through the first decoding network to obtain a first recognition result, and processing the target speech signal through the second decoding network to obtain a second recognition result; Obtaining a target recognition result corresponding to the target speech signal according to the multiple recognition results includes: A target recognition result corresponding to the target speech signal is obtained according to the first recognition result and the second recognition result.

5. The method according to claim 4, characterized in that The first decoding network includes a first sub-decoding network for processing characters corresponding to the speech signal and a second sub-decoding network for processing syllables corresponding to the speech signal; and the processing of the target speech signal by the first decoding network to obtain a first recognition result includes: The target speech signal is processed by the first sub-decoding network to obtain a first sub-recognition result, and the target speech signal is processed by the second sub-decoding network to obtain a second sub-recognition result, and the first sub-recognition result and the second sub-recognition result are used as the first recognition result.

6. The method according to claim 5, characterized in that The second decoding network includes a third sub-decoding network for processing a syllable corresponding to the speech signal; and processing the target speech signal through the second decoding network to obtain a second recognition result includes: The target speech signal is processed by the third sub-decoding network to obtain the second recognition result.

7. The method according to claim 6, characterized in that Obtaining a target recognition result corresponding to the target speech signal according to the first recognition result and the second recognition result includes: obtaining a first acoustic score for evaluating the second sub-recognition result, and obtaining a second acoustic score for evaluating the second recognition result; When the second acoustic score is higher than or equal to the first acoustic score, using the second recognition result as a target recognition result corresponding to the target speech signal; When the second acoustic score is lower than the first acoustic score, the first sub-recognition result is used as the target recognition result corresponding to the target speech signal.

8. The method according to claim 4, characterized in that The method further comprises: After updating the first decoding network, and when it is determined that a subnetwork in a second decoding network in the first server and the subnetwork in the second server are different, compiling the second decoding network by the second server to obtain a compiled second decoding network; The compiled second decoding network is pushed to the first server through the second server, and covers the second decoding network in the first server.

9. The method according to claim 8, characterized in that The second decoding network is compiled in the following manner: Obtain the voice sample corresponding to the target user; Processing the speech sample into a label sequence; The label sequence is input into a personalized decoding network to compile a sub-network in the personalized decoding network to obtain a compiled second decoding network.

10. A speech recognition device, characterized in that: The device comprises: An acquisition module, used to acquire a target voice signal of a target user; A recognition module, configured to simultaneously process the target speech signal through multiple decoding networks to obtain multiple recognition results; The acquisition module is used to obtain a target recognition result corresponding to the target speech signal based on the multiple recognition results.

11. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to implement the steps of the method according to any one of claims 1 to 9 when executing the instructions.

12. A vehicle, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to implement the steps of the method according to any one of claims 1 to 9 when executing the instructions.

13. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

14. A computer program product, characterized in that The invention comprises a computer program, which implements the steps of the method according to any one of claims 1 to 9 when the computer program is executed by a processor.

Images