CN117392996A - Target voice recognition method, device, electronic equipment and storage medium - Google Patents
Target voice recognition method, device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN117392996A CN117392996A CN202311507601.7A CN202311507601A CN117392996A CN 117392996 A CN117392996 A CN 117392996A CN 202311507601 A CN202311507601 A CN 202311507601A CN 117392996 A CN117392996 A CN 117392996A
- Authority
- CN
- China
- Prior art keywords
- speaker
- characterization
- fusion
- data
- voice
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000012512 characterization method Methods 0.000 claims abstract description 140
- 230000004927 fusion Effects 0.000 claims abstract description 124
- 108010076504 Protein Sorting Signals Proteins 0.000 claims abstract description 50
- 239000011159 matrix material Substances 0.000 claims abstract description 36
- 238000012216 screening Methods 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims description 54
- 238000012545 processing Methods 0.000 claims description 41
- 238000007499 fusion processing Methods 0.000 claims description 35
- 238000012549 training Methods 0.000 claims description 34
- 238000013507 mapping Methods 0.000 claims description 32
- 238000000605 extraction Methods 0.000 claims description 29
- 230000004913 activation Effects 0.000 claims description 24
- 210000004556 brain Anatomy 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 8
- 239000000203 mixture Substances 0.000 claims 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 6
- 238000004891 communication Methods 0.000 description 8
- 238000000926 separation method Methods 0.000 description 8
- 230000036541 health Effects 0.000 description 5
- 238000007726 management method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 3
- 108700019146 Transgenes Proteins 0.000 description 2
- 230000007177 brain activity Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0272—Voice signal separating
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0272—Voice signal separating
- G10L21/0308—Voice signal separating characterised by the type of parameter measurement, e.g. correlation techniques, zero crossing techniques or predictive techniques
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention relates to artificial intelligence, and discloses a target voice recognition method, which comprises the following steps: feature fusion is carried out on speaking voice data and an electroencephalogram signal sequence corresponding to the first speaker and the second speaker, and characterization fusion data corresponding to the first speaker and the second speaker respectively are obtained; respectively carrying out signal fusion on characterization fusion data corresponding to a first speaker and a second speaker by using a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization addition on the first dimension characterization and the second dimension characterization to obtain final characterization data; and inputting the final characterization data into a voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on the loss function and the mixed mask matrix, and taking the voice of the target speaker as the target voice. The invention also provides a target voice recognition device, electronic equipment and a storage medium. The invention can improve the accuracy of target voice recognition.
Description
Technical Field
The present invention relates to the field of artificial intelligence, and in particular, to a target voice recognition method, apparatus, electronic device, and storage medium.
Background
With the continuous progress of social and economic development, intelligent medical systems are also continuously upgraded in the field of medical health, and a scene of multi-person conference is often developed in the intelligent medical systems, so that more medical related problems are determined and solved mainly through the multi-person conference. In a multi-person conference scenario, when a plurality of different persons speak, and when different persons speak at the same time, a target speaker needs to be selected and used as a data base for subsequent voice recognition and conference recording according to the voice of the target speaker, but in the prior art, the voice of the target speaker cannot be accurately recognized, so that a more accurate target voice recognition method needs to be proposed.
Disclosure of Invention
The invention provides a target voice recognition method, a target voice recognition device, electronic equipment and a storage medium, and mainly aims to improve accuracy of target voice recognition.
In order to achieve the above object, the present invention provides a target speech recognition method, including:
Respectively acquiring speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and performing feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data respectively corresponding to the first speaker and the second speaker;
respectively carrying out signal fusion processing on the characterization fusion data corresponding to the first speaker and the second speaker by using a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization addition processing on the first dimension characterization and the second dimension characterization to obtain final characterization data;
and inputting the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and taking the voice of the target speaker as a target voice.
Optionally, the feature fusion processing is performed on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data corresponding to the first speaker and the second speaker respectively, including:
Acquiring a preset feature extraction module and a feature fusion module, and respectively carrying out feature extraction processing on speaking voice data and an electroencephalogram signal sequence of the first speaker by utilizing the feature extraction module to obtain a first voice feature and a first electroencephalogram feature;
the feature fusion module is utilized to fuse the first voice feature and the first electroencephalogram feature to obtain characterization fusion data corresponding to the first speaker;
the feature extraction module is used for carrying out feature extraction processing on the speaking voice data and the electroencephalogram signal sequence of the second speaker respectively to obtain a second voice feature and a second electroencephalogram feature;
and carrying out fusion processing on the second voice feature and the second electroencephalogram feature by utilizing the feature fusion module to obtain characterization fusion data corresponding to the second speaker.
Optionally, the inputting the final characterization data into a preset speech enhancement model to obtain a mixed mask matrix includes:
performing first linear mapping processing on the final characterization data by using a first linear mapping layer in a preset voice enhancement model to obtain first linear mapping data;
performing data conversion processing on the first linear mapping data according to two layers of transformers in the voice enhancement model to obtain conversion data;
And inputting the converted data into a second linear mapping layer in a preset voice enhancement model to perform second linear mapping processing to obtain a mixed mask matrix.
Optionally, the screening the target speaker from the first speaker and the second speaker based on the pre-constructed loss function and the mixed mask matrix includes:
calculating an activation value corresponding to the mixed mask matrix according to a preset activation function;
performing parameter optimization on the pre-trained signal fusion model by using the activation value and the pre-constructed loss function to obtain an optimized fusion model;
and inputting the first speaker and the second speaker into the optimized fusion model to obtain a target speaker.
Optionally, before performing parameter optimization on the pre-trained signal fusion model by using the activation value and the pre-constructed loss function to obtain an optimized fusion model, the method further includes:
acquiring a corresponding speaker according to the activation value, and constructing a loss function corresponding to the speaker;
and carrying out summation processing on the loss functions corresponding to the plurality of speakers to obtain a pre-constructed loss function.
Optionally, before the signal fusion processing is performed on the characterization fusion data corresponding to the first speaker and the second speaker by using the pre-trained signal fusion model, the method further includes:
Acquiring a training voice sequence and a training brain electrical signal corresponding to the training voice sequence, wherein the training brain electrical signal is an alignment signal parallel to the training voice sequence;
and performing iterative training on a preset signal model by using the training voice sequence and the training electroencephalogram signal to obtain a trained signal fusion model.
Optionally, before the step of acquiring the speaking voice data and the electroencephalogram signal sequences corresponding to the first speaker and the second speaker respectively, the method further includes:
acquiring a preset conference scene and identifying mixed voice data in the conference scene;
extracting speaking voice data corresponding to a first speaker and a second speaker from the mixed voice data;
and extracting the electroencephalogram signal sequences corresponding to the first speaker and the second speaker.
In order to solve the above problems, the present invention also provides a target voice recognition apparatus, the apparatus comprising:
the feature fusion module is used for respectively acquiring speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and carrying out feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data corresponding to the first speaker and the second speaker respectively;
The characterization adding module is used for respectively carrying out signal fusion processing on characterization fusion data corresponding to the first speaker and the second speaker by utilizing a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization adding processing on the first dimension characterization and the second dimension characterization to obtain final characterization data;
and the voice enhancement module is used for inputting the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and taking the voice of the target speaker as a target voice.
In order to solve the above-mentioned problems, the present invention also provides an electronic apparatus including:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein,
the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the target speech recognition method described above.
In order to solve the above-described problems, the present invention also provides a storage medium having stored therein at least one computer program that is executed by a processor in an electronic device to implement the above-described target voice recognition method.
In the embodiment of the invention, the feature extraction and feature fusion are carried out by combining the speaking voice data and the electroencephalogram signal sequences of different speakers, and the fusion characterization learned from the characterization fusion data by using the pre-trained signal fusion model is used in a voice enhancement module for voice separation, so that the voice enhancement and voice separation of a target speaker are realized. Therefore, the target voice recognition method, the target voice recognition device, the electronic equipment and the storage medium can solve the problem of low accuracy of improving target voice recognition.
Drawings
FIG. 1 is a flowchart of a target speech recognition method according to an embodiment of the present invention;
FIG. 2 is a detailed flow chart of one of the steps shown in FIG. 1;
FIG. 3 is a functional block diagram of a target speech recognition device according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an electronic device for implementing the target voice recognition method according to an embodiment of the present invention.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The embodiment of the application provides a target voice recognition method. The execution subject of the target voice recognition method includes, but is not limited to, at least one of a server, a terminal, and the like, which can be configured to execute the method provided by the embodiments of the present application. In other words, the target speech recognition method may be performed by software or hardware installed in a terminal device or a server device, and the software may be a blockchain platform. The service end includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like. The server may be an independent server, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms.
Referring to fig. 1, a flowchart of a target voice recognition method according to an embodiment of the invention is shown. In this embodiment, the target voice recognition method includes the following steps S1 to S3:
s1, respectively acquiring speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and performing feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data corresponding to the first speaker and the second speaker respectively.
In the embodiment of the invention, in the medical health field, the intelligent medical system is updated continuously, and in the intelligent medical system, a scene of a multi-person conference is often developed, and the speech data of a first speaker and the speech data of a second speaker in the multi-person conference are assumed to be acquired, wherein the first speaker and the second speaker can be doctors or patients, and can also be other speakers in the multi-person conference in other medical health fields. The Electroencephalogram signal sequence refers to an Electroencephalogram signal, and the Electroencephalogram signal sequence (EEG) is a method for recording brain activity by using an electrophysiological index.
Specifically, before the speaking voice data and the electroencephalogram signal sequences corresponding to the first speaker and the second speaker are respectively obtained, the method further includes:
Acquiring a preset conference scene and identifying mixed voice data in the conference scene;
extracting speaking voice data corresponding to a first speaker and a second speaker from the mixed voice data;
and extracting the electroencephalogram signal sequences corresponding to the first speaker and the second speaker.
In detail, the mixed voice data includes voices of a plurality of speakers in a conference scene.
Further, referring to fig. 2, the feature fusion processing is performed on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data corresponding to the first speaker and the second speaker respectively, including:
s11, acquiring a preset feature extraction module and a feature fusion module, and respectively carrying out feature extraction processing on speaking voice data and an electroencephalogram signal sequence of the first speaker by utilizing the feature extraction module to obtain a first voice feature and a first electroencephalogram feature;
s12, carrying out fusion processing on the first voice feature and the first electroencephalogram feature by utilizing the feature fusion module to obtain characterization fusion data corresponding to the first speaker;
s13, respectively carrying out feature extraction processing on the speaking voice data and the electroencephalogram signal sequence of the second speaker by utilizing the feature extraction module to obtain a second voice feature and a second electroencephalogram feature;
S14, fusion processing is carried out on the second voice feature and the second brain electrical feature by utilizing the feature fusion module, and characterization fusion data corresponding to the second speaker is obtained.
In detail, the feature extraction module Encoder is utilized to respectively perform feature extraction processing on the speaking voice data and the electroencephalogram signal sequence of the first speaker to obtain a first voice feature Z 1 And a first electroencephalogram feature Z 2 The feature fusion module is utilized to carry out the feature fusion on the first voice feature Z 1 And the first electroencephalogram feature Z 2 FuSion processing is carried out to obtain characterization FuSion data Z=FuSion (encoder (Audio), encoder (EEG) 1 ) Wherein the encoder is feature extraction and the Audio is speaking voice data of the first speaker, EEG 1 Is an electroencephalogram signal sequence of the first speaker.
The feature fusion module can be composed of a D layer, each layer is composed of a 1D convolution layer, a normalization layer and a GELU activation layer, the feature fusion module can be realized by using an attention mechanism, and the characterization fusion data corresponding to the first speaker and the second speaker respectively are low-latitude characterization.
S2, respectively carrying out signal fusion processing on the characterization fusion data corresponding to the first speaker and the second speaker by utilizing a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization addition processing on the first dimension characterization and the second dimension characterization to obtain final characterization data.
In the embodiment of the invention, the pre-trained signal fusion model is an AE-Hubert model, and the AE-Hubert model is a multi-modal voice separation model.
Specifically, before the signal fusion processing is performed on the characterization fusion data corresponding to the first speaker and the second speaker by using the pre-trained signal fusion model, the method further includes:
acquiring a training voice sequence and a training brain electrical signal corresponding to the training voice sequence, wherein the training brain electrical signal is an alignment signal parallel to the training voice sequence;
and performing iterative training on a preset signal model by using the training voice sequence and the training electroencephalogram signal to obtain a trained signal fusion model.
In detail, in the case of the speech sequence A1: setting parallel aligned electroencephalogram signal frame sequence E on basis of single encoder of T 1：T Wherein n represents the speaker, T represents the electroencephalogram signal or the mixed voice signal corresponding to the time frame, the encoder is trained on the electroencephalogram signal data of the speaker by using time, and iterative training is alternately carried out between the two encoders, so that a pre-trained audio-electroencephalogram signal Hubert model is finally obtained.
Further, signal fusion processing is carried out on the characterization fusion data corresponding to the first speaker and the second speaker respectively by utilizing a pre-trained signal fusion model, so as to obtain a first-dimension characterization O 1 And a second dimension representation O 2 Will be gotPerforming characterization addition processing on the first dimension characterization and the second dimension characterization to obtain final characterization data as O= (O) 1 +O 2 )。
In detail, the learned representation is input into a HuBert model by using a pre-trained AE-Hubert model, the two pre-trained models branch learn high-dimensional latent layer representation of voice, the representations learned by the two branch models are added to serve as input data of a voice enhancement model, and the model output is O 1 ,O 2 The output of the overall high-dimensional latent layer representation is o= (O) 1 +O 2 ) As input data of the speech enhancement module, a mixing mask corresponding to the speaker is learned.
S3, inputting the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and taking the voice of the target speaker as a target voice.
In the embodiment of the present invention, the step of inputting the final characterization data into a preset speech enhancement model to obtain a mixed mask matrix includes:
performing first linear mapping processing on the final characterization data by using a first linear mapping layer in a preset voice enhancement model to obtain first linear mapping data;
performing data conversion processing on the first linear mapping data according to two layers of transformers in the voice enhancement model to obtain conversion data;
and inputting the converted data into a second linear mapping layer in a preset voice enhancement model to perform second linear mapping processing to obtain a mixed mask matrix.
In detail, the speech enhancement model is composed of one linear mapping layer, two transgenes, one linear mapping layer, and one Sigmoid. Wherein the output dimension corresponding to the first linear mapping layer is set to 256, the input and output dimensions of the two-layer transducer are set to (256), and the input and output dimensions corresponding to the next linear mapping layer are set to (256, 512).
Specifically, the screening the target speaker from the first speaker and the second speaker based on the pre-constructed loss function and the mixed mask matrix includes:
Calculating an activation value corresponding to the mixed mask matrix according to a preset activation function;
performing parameter optimization on the pre-trained signal fusion model by using the activation value and the pre-constructed loss function to obtain an optimized fusion model;
and inputting the first speaker and the second speaker into the optimized fusion model to obtain a target speaker.
In detail, in this scheme, the preset activation function is a softmax function.
Further, the method further includes, before performing parameter optimization on the pre-trained signal fusion model by using the activation value and the pre-constructed loss function to obtain an optimized fusion model:
acquiring a corresponding speaker according to the activation value, and constructing a loss function corresponding to the speaker;
and carrying out summation processing on the loss functions corresponding to the plurality of speakers to obtain a pre-constructed loss function.
In detail, the pre-constructed loss function is
L se ＝L sp1 +L sp2
Wherein l sp1 And l sp2 Respectively representing the corresponding loss functions of the two speakers.
In the embodiment of the invention, the feature extraction and feature fusion are carried out by combining the speaking voice data and the electroencephalogram signal sequences of different speakers, and the fusion characterization learned from the characterization fusion data by using the pre-trained signal fusion model is used in a voice enhancement module for voice separation, so that the voice enhancement and voice separation of a target speaker are realized. Therefore, the target voice recognition method provided by the invention can solve the problem of low accuracy of improving target voice recognition.
Fig. 3 is a functional block diagram of a target voice recognition device according to an embodiment of the present invention.
The target speech recognition apparatus 100 of the present invention may be installed in an electronic device. Depending on the functions implemented, the target speech recognition device 100 may include a feature fusion module 101, a token addition module 102, and a speech enhancement module 103. The module of the invention, which may also be referred to as a unit, refers to a series of computer program segments, which are stored in the memory of the electronic device, capable of being executed by the processor of the electronic device and of performing a fixed function.
In the present embodiment, the functions concerning the respective modules/units are as follows:
the feature fusion module 101 is configured to obtain speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and perform feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data corresponding to the first speaker and the second speaker respectively;
the characterization adding module 102 is configured to perform signal fusion processing on characterization fusion data corresponding to the first speaker and the second speaker by using a pre-trained signal fusion model, to obtain a first dimension characterization and a second dimension characterization, and perform characterization adding processing on the first dimension characterization and the second dimension characterization, to obtain final characterization data;
The voice enhancement module 103 is configured to input the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screen a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and take the voice of the target speaker as a target voice.
In detail, the specific embodiments of the modules of the target voice recognition device 100 are as follows:
step one, respectively acquiring speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and carrying out feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data respectively corresponding to the first speaker and the second speaker.
In the embodiment of the invention, in the medical health field, the intelligent medical system is updated continuously, and in the intelligent medical system, a scene of a multi-person conference is often developed, and the speech data of a first speaker and the speech data of a second speaker in the multi-person conference are assumed to be acquired, wherein the first speaker and the second speaker can be doctors or patients, and can also be other speakers in the multi-person conference in other medical health fields. The Electroencephalogram signal sequence refers to an Electroencephalogram signal, and the Electroencephalogram signal sequence (EEG) is a method for recording brain activity by using an electrophysiological index.
Specifically, before the speaking voice data and the electroencephalogram signal sequences corresponding to the first speaker and the second speaker are respectively obtained, the method further includes:
acquiring a preset conference scene and identifying mixed voice data in the conference scene;
extracting speaking voice data corresponding to a first speaker and a second speaker from the mixed voice data;
and extracting the electroencephalogram signal sequences corresponding to the first speaker and the second speaker.
In detail, the mixed voice data includes voices of a plurality of speakers in a conference scene.
Further, the feature fusion processing is performed on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data corresponding to the first speaker and the second speaker respectively, including:
acquiring a preset feature extraction module and a feature fusion module, and respectively carrying out feature extraction processing on speaking voice data and an electroencephalogram signal sequence of the first speaker by utilizing the feature extraction module to obtain a first voice feature and a first electroencephalogram feature;
the feature fusion module is utilized to fuse the first voice feature and the first electroencephalogram feature to obtain characterization fusion data corresponding to the first speaker;
The feature extraction module is used for carrying out feature extraction processing on the speaking voice data and the electroencephalogram signal sequence of the second speaker respectively to obtain a second voice feature and a second electroencephalogram feature;
and carrying out fusion processing on the second voice feature and the second electroencephalogram feature by utilizing the feature fusion module to obtain characterization fusion data corresponding to the second speaker.
In detail, the feature extraction module Encoder is utilized to respectively perform feature extraction processing on the speaking voice data and the electroencephalogram signal sequence of the first speaker to obtain a first voice feature Z 1 And a first electroencephalogram feature Z 2 The feature fusion module is utilized to carry out the feature fusion on the first voice feature Z 1 And the first electroencephalogram feature Z 2 FuSion processing is carried out to obtain characterization FuSion data Z=FuSion (encoder (Audio), encoder (EEG) 1 ) Wherein the encoder is feature extraction and the Audio is speaking voice data of the first speaker, EEG 1 Is an electroencephalogram signal sequence of the first speaker.
The feature fusion module can be composed of a D layer, each layer is composed of a 1D convolution layer, a normalization layer and a GELU activation layer, the feature fusion module can be realized by using an attention mechanism, and the characterization fusion data corresponding to the first speaker and the second speaker respectively are low-latitude characterization.
And secondly, respectively carrying out signal fusion processing on the characterization fusion data corresponding to the first speaker and the second speaker by using a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization addition processing on the first dimension characterization and the second dimension characterization to obtain final characterization data.
In the embodiment of the invention, the pre-trained signal fusion model is an AE-Hubert model, and the AE-Hubert model is a multi-modal voice separation model.
Specifically, before the signal fusion processing is performed on the characterization fusion data corresponding to the first speaker and the second speaker by using the pre-trained signal fusion model, the method further includes:
acquiring a training voice sequence and a training brain electrical signal corresponding to the training voice sequence, wherein the training brain electrical signal is an alignment signal parallel to the training voice sequence;
and performing iterative training on a preset signal model by using the training voice sequence and the training electroencephalogram signal to obtain a trained signal fusion model.
In detail, in the case of the speech sequence A1: setting parallel aligned electroencephalogram signal frame sequence E on basis of single encoder of T 1：T Wherein n represents the speaker, T represents the electroencephalogram signal or the mixed voice signal corresponding to the time frame, the encoder is trained on the electroencephalogram signal data of the speaker by using time, and iterative training is alternately carried out between the two encoders, so that a pre-trained audio-electroencephalogram signal Hubert model is finally obtained.
Further, signal fusion processing is carried out on the characterization fusion data corresponding to the first speaker and the second speaker respectively by utilizing a pre-trained signal fusion model, so as to obtain a first-dimension characterization O 1 And a second dimension representation O 2 Performing characterization addition processing on the first dimension characterization and the second dimension characterization to obtain final characterization data which is O= (O) 1 +O 2 )。
In detail, the learned representation is input into a HuBert model by using a pre-trained AE-Hubert model, the two pre-trained models branch learn high-dimensional latent layer representation of voice, the representations learned by the two branch models are added to serve as input data of a voice enhancement model, and the model output is O 1 ,O 2 The output of the overall high-dimensional latent layer representation is o= (O) 1 +O 2 ) As input data of the speech enhancement module, a mixing mask corresponding to the speaker is learned.
Inputting the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and taking the voice of the target speaker as a target voice.
In the embodiment of the present invention, the step of inputting the final characterization data into a preset speech enhancement model to obtain a mixed mask matrix includes:
performing first linear mapping processing on the final characterization data by using a first linear mapping layer in a preset voice enhancement model to obtain first linear mapping data;
performing data conversion processing on the first linear mapping data according to two layers of transformers in the voice enhancement model to obtain conversion data;
and inputting the converted data into a second linear mapping layer in a preset voice enhancement model to perform second linear mapping processing to obtain a mixed mask matrix.
In detail, the speech enhancement model is composed of one linear mapping layer, two transgenes, one linear mapping layer, and one Sigmoid. Wherein the output dimension corresponding to the first linear mapping layer is set to 256, the input and output dimensions of the two-layer transducer are set to (256), and the input and output dimensions corresponding to the next linear mapping layer are set to (256, 512).
Specifically, the screening the target speaker from the first speaker and the second speaker based on the pre-constructed loss function and the mixed mask matrix includes:
calculating an activation value corresponding to the mixed mask matrix according to a preset activation function;
performing parameter optimization on the pre-trained signal fusion model by using the activation value and the pre-constructed loss function to obtain an optimized fusion model;
and inputting the first speaker and the second speaker into the optimized fusion model to obtain a target speaker.
In detail, in this scheme, the preset activation function is a softmax function.
Further, the method further includes, before performing parameter optimization on the pre-trained signal fusion model by using the activation value and the pre-constructed loss function to obtain an optimized fusion model:
acquiring a corresponding speaker according to the activation value, and constructing a loss function corresponding to the speaker;
and carrying out summation processing on the loss functions corresponding to the plurality of speakers to obtain a pre-constructed loss function.
In detail, the pre-constructed loss function is
L se ＝L sp1 +L sp2
Wherein l sp1 And l sp2 Respectively representing the corresponding loss functions of the two speakers.
In the embodiment of the invention, the feature extraction and feature fusion are carried out by combining the speaking voice data and the electroencephalogram signal sequences of different speakers, and the fusion characterization learned from the characterization fusion data by using the pre-trained signal fusion model is used in a voice enhancement module for voice separation, so that the voice enhancement and voice separation of a target speaker are realized. Therefore, the target voice recognition device provided by the invention can solve the problem of low accuracy of improving target voice recognition.
Fig. 4 is a schematic structural diagram of an electronic device for implementing a target voice recognition method according to an embodiment of the present invention.
The electronic device 1 may comprise a processor 10, a memory 11, a communication bus 12 and a communication interface 13, and may further comprise a computer program, such as a target speech recognition program, stored in the memory 11 and executable on the processor 10.
The processor 10 may be formed by an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or may be formed by a plurality of integrated circuits packaged with the same function or different functions, including one or more central processing units (Central Processing Unit, CPU), a microprocessor, a digital processing chip, a graphics processor, a combination of various control chips, and so on. The processor 10 is a Control Unit (Control Unit) of the electronic device, connects various components of the entire electronic device using various interfaces and lines, and executes various functions of the electronic device and processes data by running or executing programs or modules (e.g., executing a target voice recognition program, etc.) stored in the memory 11, and calling data stored in the memory 11.
The memory 11 includes at least one type of readable storage medium including flash memory, a removable hard disk, a multimedia card, a card type memory (e.g., SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. The memory 11 may in some embodiments be an internal storage unit of the electronic device, such as a mobile hard disk of the electronic device. The memory 11 may in other embodiments also be an external storage device of the electronic device, such as a plug-in mobile hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card) or the like, which are provided on the electronic device. Further, the memory 11 may also include both an internal storage unit and an external storage device of the electronic device. The memory 11 may be used not only for storing application software installed in an electronic device and various types of data, such as codes of a target voice recognition program, but also for temporarily storing data that has been output or is to be output.
The communication bus 12 may be a peripheral component interconnect standard (Peripheral Component Interconnect, PCI) bus, or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, among others. The bus may be classified as an address bus, a data bus, a control bus, etc. The bus is arranged to enable a connection communication between the memory 11 and at least one processor 10 etc.
The communication interface 13 is used for communication between the electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and/or a wireless interface (e.g., WI-FI interface, bluetooth interface, etc.), typically used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a Display (Display), an input unit such as a Keyboard (Keyboard), or alternatively a standard wired interface, a wireless interface. Alternatively, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, or the like. The display may also be referred to as a display screen or display unit, as appropriate, for displaying information processed in the electronic device and for displaying a visual user interface.
Fig. 4 shows only an electronic device with components, it being understood by a person skilled in the art that the structure shown in fig. 4 does not constitute a limitation of the electronic device 1, and may comprise fewer or more components than shown, or may combine certain components, or may be arranged in different components.
For example, although not shown, the electronic device may further include a power source (such as a battery) for supplying power to the respective components, and preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that functions of charge management, discharge management, power consumption management, and the like are implemented through the power management device. The power supply may also include one or more of any of a direct current or alternating current power supply, recharging device, power failure detection circuit, power converter or inverter, power status indicator, etc. The electronic device may further include various sensors, bluetooth modules, wi-Fi modules, etc., which are not described herein.
It should be understood that the embodiments described are for illustrative purposes only and are not limited to this configuration in the scope of the patent application.
The target speech recognition program stored in the memory 11 of the electronic device 1 is a combination of instructions which, when run in the processor 10, can implement:
respectively acquiring speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and performing feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data respectively corresponding to the first speaker and the second speaker;
Respectively carrying out signal fusion processing on the characterization fusion data corresponding to the first speaker and the second speaker by using a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization addition processing on the first dimension characterization and the second dimension characterization to obtain final characterization data;
and inputting the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and taking the voice of the target speaker as a target voice.
In particular, the specific implementation method of the above instructions by the processor 10 may refer to the description of the relevant steps in the corresponding embodiment of the drawings, which is not repeated herein.
Further, the modules/units integrated in the electronic device 1 may be stored in a storage medium if implemented in the form of software functional units and sold or used as separate products. The storage medium may be volatile or nonvolatile. For example, the computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM).
The present invention also provides a storage medium storing a computer program which, when executed by a processor of an electronic device, can implement:
respectively acquiring speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and performing feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data respectively corresponding to the first speaker and the second speaker;
respectively carrying out signal fusion processing on the characterization fusion data corresponding to the first speaker and the second speaker by using a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization addition processing on the first dimension characterization and the second dimension characterization to obtain final characterization data;
and inputting the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and taking the voice of the target speaker as a target voice.
In the several embodiments provided in the present invention, it should be understood that the disclosed apparatus, device and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a logical function division, and there may be other manners of division when actually implemented.
The modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, may be located in one place, or may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional module in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units can be realized in a form of hardware or a form of hardware and a form of software functional modules.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned.
The blockchain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm and the like. The Blockchain (Blockchain), which is essentially a decentralised database, is a string of data blocks that are generated by cryptographic means in association, each data block containing a batch of information of network transactions for verifying the validity of the information (anti-counterfeiting) and generating the next block. The blockchain may include a blockchain underlying platform, a platform product services layer, an application services layer, and the like.
The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results.
Furthermore, it is evident that the word "comprising" does not exclude other elements or steps, and that the singular does not exclude a plurality. A plurality of units or means recited in the system claims can also be implemented by means of software or hardware by means of one unit or means. The terms first, second, etc. are used to denote a name, but not any particular order.
Finally, it should be noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.
Claims (10)
1. A method of target speech recognition, the method comprising:
respectively acquiring speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and performing feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data respectively corresponding to the first speaker and the second speaker;
respectively carrying out signal fusion processing on the characterization fusion data corresponding to the first speaker and the second speaker by using a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization addition processing on the first dimension characterization and the second dimension characterization to obtain final characterization data;
And inputting the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and taking the voice of the target speaker as a target voice.
2. The method of claim 1, wherein the feature fusion processing is performed on the speaker voice data and the electroencephalogram signal sequence to obtain characterization fusion data corresponding to the first speaker and the second speaker, respectively, including:
acquiring a preset feature extraction module and a feature fusion module, and respectively carrying out feature extraction processing on speaking voice data and an electroencephalogram signal sequence of the first speaker by utilizing the feature extraction module to obtain a first voice feature and a first electroencephalogram feature;
the feature fusion module is utilized to fuse the first voice feature and the first electroencephalogram feature to obtain characterization fusion data corresponding to the first speaker;
the feature extraction module is used for carrying out feature extraction processing on the speaking voice data and the electroencephalogram signal sequence of the second speaker respectively to obtain a second voice feature and a second electroencephalogram feature;
And carrying out fusion processing on the second voice feature and the second electroencephalogram feature by utilizing the feature fusion module to obtain characterization fusion data corresponding to the second speaker.
3. The method of claim 1, wherein the inputting the final characterization data into a predetermined speech enhancement model to obtain a mixture mask matrix comprises:
performing first linear mapping processing on the final characterization data by using a first linear mapping layer in a preset voice enhancement model to obtain first linear mapping data;
performing data conversion processing on the first linear mapping data according to two layers of transformers in the voice enhancement model to obtain conversion data;
and inputting the converted data into a second linear mapping layer in a preset voice enhancement model to perform second linear mapping processing to obtain a mixed mask matrix.
4. The target speech recognition method of claim 1, wherein the screening target speakers from the first speaker and the second speaker based on the pre-constructed loss function and the mixture mask matrix comprises:
calculating an activation value corresponding to the mixed mask matrix according to a preset activation function;
Performing parameter optimization on the pre-trained signal fusion model by using the activation value and the pre-constructed loss function to obtain an optimized fusion model;
and inputting the first speaker and the second speaker into the optimized fusion model to obtain a target speaker.
5. The method for target speech recognition according to claim 1, wherein the method further comprises, before performing parameter optimization on a pre-trained signal fusion model by using the activation value and a pre-constructed loss function to obtain an optimized fusion model:
acquiring a corresponding speaker according to the activation value, and constructing a loss function corresponding to the speaker;
and carrying out summation processing on the loss functions corresponding to the plurality of speakers to obtain a pre-constructed loss function.
6. The method of claim 1, wherein before the signal fusion processing is performed on the token fusion data corresponding to the first speaker and the second speaker by using a pre-trained signal fusion model, the method further comprises:
acquiring a training voice sequence and a training brain electrical signal corresponding to the training voice sequence, wherein the training brain electrical signal is an alignment signal parallel to the training voice sequence;
And performing iterative training on a preset signal model by using the training voice sequence and the training electroencephalogram signal to obtain a trained signal fusion model.
7. The method of claim 1, wherein before the step of obtaining the speech data and the electroencephalogram signal sequences corresponding to the first speaker and the second speaker, respectively, the method further comprises:
acquiring a preset conference scene and identifying mixed voice data in the conference scene;
extracting speaking voice data corresponding to a first speaker and a second speaker from the mixed voice data;
and extracting the electroencephalogram signal sequences corresponding to the first speaker and the second speaker.
8. A target speech recognition device, the device comprising:
the feature fusion module is used for respectively acquiring speaking voice data and an electroencephalogram signal sequence corresponding to a first speaker and a second speaker, and carrying out feature fusion processing on the speaking voice data and the electroencephalogram signal sequence to obtain characterization fusion data corresponding to the first speaker and the second speaker respectively;
the characterization adding module is used for respectively carrying out signal fusion processing on characterization fusion data corresponding to the first speaker and the second speaker by utilizing a pre-trained signal fusion model to obtain a first dimension characterization and a second dimension characterization, and carrying out characterization adding processing on the first dimension characterization and the second dimension characterization to obtain final characterization data;
And the voice enhancement module is used for inputting the final characterization data into a preset voice enhancement model to obtain a mixed mask matrix, screening a target speaker from the first speaker and the second speaker based on a pre-constructed loss function and the mixed mask matrix, and taking the voice of the target speaker as a target voice.
9. An electronic device, the electronic device comprising:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein,
the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the target speech recognition method according to any one of claims 1 to 7.
10. A storage medium storing a computer program which, when executed by a processor, implements the target speech recognition method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311507601.7A CN117392996A (en) | 2023-11-13 | 2023-11-13 | Target voice recognition method, device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311507601.7A CN117392996A (en) | 2023-11-13 | 2023-11-13 | Target voice recognition method, device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117392996A true CN117392996A (en) | 2024-01-12 |
Family
ID=89466578
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311507601.7A Pending CN117392996A (en) | 2023-11-13 | 2023-11-13 | Target voice recognition method, device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117392996A (en) |
-
2023
- 2023-11-13 CN CN202311507601.7A patent/CN117392996A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107767929A (en) | CRF filling method, device, electronic equipment and storage medium | |
| CN114822812A (en) | Character dialogue simulation method, device, equipment and storage medium | |
| CN113064994A (en) | Conference quality evaluation method, device, equipment and storage medium | |
| CN114491047A (en) | Multi-label text classification method and device, electronic equipment and storage medium | |
| CN113627160B (en) | Text error correction method and device, electronic equipment and storage medium | |
| CN115600644A (en) | Multitasking method and device, electronic equipment and storage medium | |
| CN116630457A (en) | Training method and device for picture generation model, electronic equipment and storage medium | |
| CN114913371A (en) | Multitask learning model training method and device, electronic equipment and storage medium | |
| CN114780701A (en) | Automatic question-answer matching method, device, computer equipment and storage medium | |
| CN116383766A (en) | Auxiliary diagnosis method, device, equipment and storage medium based on multi-mode data | |
| CN115409041B (en) | Unstructured data extraction method, device, equipment and storage medium | |
| CN116702776A (en) | Multi-task semantic division method, device, equipment and medium based on cross-Chinese and western medicine | |
| CN116630712A (en) | Information classification method and device based on modal combination, electronic equipment and medium | |
| CN116705304A (en) | Multi-mode task processing method, device, equipment and medium based on image text | |
| CN116719904A (en) | Information query method, device, equipment and storage medium based on image-text combination | |
| CN114548114B (en) | Text emotion recognition method, device, equipment and storage medium | |
| CN117392996A (en) | Target voice recognition method, device, electronic equipment and storage medium | |
| CN114610855A (en) | Dialog reply generation method and device, electronic equipment and storage medium | |
| CN112560721B (en) | Non-perception model switching method and device, electronic equipment and storage medium | |
| CN113921097A (en) | Medical data integration method and device, electronic equipment and storage medium | |
| CN113157677A (en) | Data filtering method and device based on trust behaviors | |
| CN115221875B (en) | Word weight generation method, device, electronic equipment and storage medium | |
| CN116631608A (en) | Method, device, equipment and storage medium for identifying inquiry symptoms | |
| CN114742423B (en) | Random grouping method, device, equipment and medium | |
| CN112232408A (en) | Target recommendation method and device, electronic equipment and computer-readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |