CN109714692A - Noise reduction method based on personal data and artificial neural network - Google Patents
Noise reduction method based on personal data and artificial neural network Download PDFInfo
- Publication number
- CN109714692A CN109714692A CN201811602590.XA CN201811602590A CN109714692A CN 109714692 A CN109714692 A CN 109714692A CN 201811602590 A CN201811602590 A CN 201811602590A CN 109714692 A CN109714692 A CN 109714692A
- Authority
- CN
- China
- Prior art keywords
- neural network
- noise reduction
- layer
- hearing aid
- data
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 68
- 230000009467 reduction Effects 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 22
- 230000008569 process Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 5
- 238000009825 accumulation Methods 0.000 claims description 4
- 230000005284 excitation Effects 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 2
- 230000008030 elimination Effects 0.000 abstract description 2
- 238000003379 elimination reaction Methods 0.000 abstract description 2
- 238000004519 manufacturing process Methods 0.000 abstract 1
- 238000005457 optimization Methods 0.000 abstract 1
- 206010011878 Deafness Diseases 0.000 description 15
- 230000010370 hearing loss Effects 0.000 description 10
- 231100000888 hearing loss Toxicity 0.000 description 10
- 208000016354 hearing loss disease Diseases 0.000 description 10
- 230000032683 aging Effects 0.000 description 3
- 230000007423 decrease Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 208000024827 Alzheimer disease Diseases 0.000 description 1
- 208000017667 Chronic Disease Diseases 0.000 description 1
- 241001062009 Indigofera Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000747 cardiac effect Effects 0.000 description 1
- 230000003920 cognitive function Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005802 health problem Effects 0.000 description 1
- 230000013011 mating Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
Landscapes
- Circuit For Audible Band Transducer (AREA)
Abstract
本发明属于助听器噪声消除技术领域，为提出贴合用户个人的成熟降噪算法。为此，本发明采取的技术方案是，基于个人数据与人工神经网络的助听器降噪方法，通过分析助听器使用者接收到的语音信号，识别并储存有话段，利用储存的足够数量的语音信息对BP人工神经网络进行训练。通过训练对神经网络各层的权值与阈值的优化调整，以得到成熟的BP人工神经网络，来修正常规的语音降噪公式，进而令助听器使用者得到针对其个人优化过的更加清晰的降噪后语音。本发明主要应用于助听器设计制造场合。
The invention belongs to the technical field of noise elimination of hearing aids, and aims to propose a mature noise reduction algorithm suitable for individual users. To this end, the technical solution adopted by the present invention is that the noise reduction method of hearing aids based on personal data and artificial neural network, by analyzing the speech signals received by the hearing aid users, identify and store the speech segments, and use the stored enough voice information. The BP artificial neural network is trained. By training the optimization and adjustment of the weights and thresholds of each layer of the neural network, a mature BP artificial neural network can be obtained to modify the conventional speech noise reduction formula, so that the hearing aid users can obtain a clearer noise reduction optimized for their individual. post-noise speech. The present invention is mainly applied to the occasion of hearing aid design and manufacture.
Description
Technical field
The invention belongs to hearing aid noise cancelling alorithm design fields, are related to one kind based on user's personal data and BP people
The digital deaf-aid noise reduction algorithm positioned at Portable intelligent terminal that artificial neural networks are realized.
Background technique
Currently, world's Aging Problem is got worse, Chinese Aged ratio is also increasing year by year, and China is made to become the world
Upper aging of population one of country with the fastest developing speed.The middle of this century is expected, China there will be nearly 500,000,000 population to reach the world
The elderly's standard of health organization.
Hearing loss is then one of most common several chronic diseases of the elderly.According to the World Health Organization in 2015
Statistics, the 5% of world population suffers from disabling hearing loss, and most of among these is the elderly.Hearing loss may
Cause cognitive function decline, depression, the more serious psychological problems such as happiness reduction.Even studies have shown that: it is light, in, again
The old man of hearing loss is spent, the illness rate of Alzheimer disease is 2 times, 3 times and 5 times of the normal old man of hearing respectively.It can be seen that
Hearing loss is the serious health problems urgently to be resolved occurred with whole world aging.
Hearing aid can play great booster action for the hearing improved of hearing loss person.Hearing aid can not only mention
The social contact ability and personal lifestyle quality of high hearing loss person, and the generation of disease can be managed to avoid severe cardiac.Therefore,
With the raising of people's economic level and the continuous pursuit to high-quality life, the ratio of hearing aid is used in hearing loss person
It is being continuously improved.In the use process of hearing aid, voice signal inevitably will receive the influence of ambient noise, make
Obtain the identification degree decline of voice signal.And since hearing loss person is more insensitive to sound, hearing aid usage experience meeting
It is influenced by compared to normal person is bigger.Therefore the voice de-noising function of hearing aid is most important.The main mesh of voice de-noising
Be that elimination noise improves signal-to-noise ratio (signal to noise ratio, SNR) as much as possible to improve voice quality, and
Thus loss person is improved one's hearing to the identification of voice.
Since the daily local environment of each hearing aid user and Communicator are different, therefore voice signal and noise are believed
Number the characteristics of be also the noise reduction effect that has very big difference, therefore reached together using conventional noise reduction algorithm not ten sub-argument
Think.So we should be directed to noise reduction formula according to the surroundings of each user and the exclusive feature of Communicator
Property adjust automatically, to achieve the purpose that accurate noise reduction.Unlike regular speech noise-reduction method, ideal voice de-noising algorithm
The individual differences for user should be focused on, this can greatly improve the hearing aid usage experience of target hearing loss person.
Summary of the invention
In order to overcome the deficiencies of the prior art, the present invention is directed to propose it is a kind of based on BP (back propagation, reverse biography
Broadcast) the digital deaf-aid noise reduction algorithm of artificial neural network.The parameter that the noise reduction algorithm obtains a large amount of random voice data
As initial parameter, the voice signal that user is received constantly train as training data, to BP artificial neural network,
Obtain the mature noise reduction algorithm of fitting individual subscriber.For this reason, the technical scheme adopted by the present invention is that based on personal data and manually
The Noise reduction method of neural network, the voice signal received by analyzing hearing aid user, identifies and stores words
Section, is trained BP artificial neural network using the sufficient amount of voice messaging of storage.It is each to neural network by training
The weight of layer and optimizing and revising for threshold value, to obtain mature BP artificial neural network, to correct conventional voice de-noising formula,
And then hearing aid user is enabled to obtain voice after the noise reduction being more clear optimized for its people.
Further, noise reduction algorithm is set to the operation of hearing aid end, neural network algorithm is set to smart machine fortune
Row, smart machine are wirelessly connect with hearing aid.User can independently set voice accumulation, whenever the voice of accumulation
When amount reaches user's setting value, hearing aid sends smart machine by bluetooth module for these voice data automatically, and by intelligence
Energy equipment is stored and is handled, and smart machine is trained artificial neural network using these voice data, and will train
Parameter afterwards is sent to hearing aid, is used with adjusting noise reduction algorithm fitting individual subscriber.User can with unrestricted choice by voice,
Neural network parameter data upload to cloud or download from cloud, to realize the data sharing between multiple smart machines.
The topological structure of BP artificial neural network includes: input layer, hidden layer and output layer three-decker, due to input
Voice signal needs to extract phonetic feature by mel cepstrum coefficients method, and voice signal is by becoming 24 after Meier filter group
Dimensional feature signal, therefore input layer includes 24 neurodes, initial hidden layer includes 25 neurodes, and number of nodes
It can be adjusted with the variation of training data, output layer then includes 30 neurodes, and is exported for multichannel, corresponding noise reduction
Subtracting coefficient, gain compensation factor parameter, the company of input layer and hidden layer, hidden layer and output layer are crossed under different frequency in algorithm
Connecing weight is respectively ωij,ωjk, hidden layer threshold value a, output layer threshold value is b, ωij、ωjk, a, b these parameters initial value
The neural network practiced from training on trial, i.e., the neural network obtained using a large amount of random voice data training, training process are used
BP artificial neural network commonly trains formula, and input layer, hidden layer, output layer number of nodes are denoted as m, l, n in formula:
Step 1: hidden layer, which is calculated, using general hidden layer excitation function exports, H in formulajWith ajRespectively correspond the hidden of j-th of node
Containing layer output and hidden layer threshold value
Step 2: calculating output layer error, O in formulakWith ekRespectively correspond the output and output error of k-th of node, YkFor
Desired output from training data
ek=Yk-Ok
K=1,2 ..., l
Step 3: updating weight and threshold value, η is learning rate, b in formulakFor the output layer threshold value of k-th of node
ωjk=ωjk+ηHjek
bk=bk+ek
Continuous iteration is carried out using above-mentioned formula, the number of iterations until reaching setting.
The training process of neural network used is as follows: when user has accumulated enough voices in hearing aid use process
After data, neural network algorithm can start to run in smart machine end, and when operation can independently set BP artificial neural network first
The structure of network, the number of nodes including the number of plies Yu each layer, and by the parameter settings such as the threshold value of each layer, weight be program give it is initial
Value, connect it is lower by data by Meier filter group and start train artificial neural network, after reaching the number of iterations of setting, journey
Sequence can send output parameter to hearing aid and thus adjust noise reduction algorithm, and hereafter hearing aid enters the routine use stage, until
After having accumulated enough new speech data, algorithm can select new data and legacy data to input neural network according to a certain percentage,
It is trained again.
The features of the present invention and beneficial effect are:
Compared to other Noise reduction algorithms, the digital deaf-aid based on BP artificial neural network that the present invention uses drops
Algorithm of making an uproar is more mature.And since the algorithm is intended to be adjusted for users personal data, noise reduction precision also can be big
It is big to improve, can preferably optimize the usage experience of user, allow users to receive be more clear, comfortable voice.
Detailed description of the invention:
Fig. 1 is the topology diagram of BP artificial neural network used in the present invention.
Fig. 2 is the training flow chart of neural network used in the present invention.
Fig. 3 is hearing aid in the present invention, smart machine and the relational graph in cloud.
Specific embodiment
The invention belongs to hearing aid noise cancelling alorithm design fields, are related to one kind based on user's personal data and BP people
The digital deaf-aid noise reduction algorithm positioned at Portable intelligent terminal that artificial neural networks are realized.The invention is made by analyzing hearing aid
The voice signal that user receives identifies and stores words section, manually refreshing to BP using the sufficient amount of voice messaging of storage
It is trained through network.By training to the weight of each layer of neural network and optimizing and revising for threshold value, to obtain mature BP people
Artificial neural networks to correct conventional voice de-noising formula, and then enable hearing aid user obtain being directed to what its people optimized
Voice after the noise reduction being more clear.
The present invention provides a kind of digital deaf-aid noise reduction algorithm based on BP artificial neural network.The noise reduction algorithm will be a large amount of
The parameter that random voice data obtains is as initial parameter, and the voice signal that user is received is as training data, to BP
Artificial neural network constantly train, and obtains the mature noise reduction algorithm of fitting individual subscriber.
BP artificial neural network is a kind of multilayer feedforward neural network according to error backpropagation algorithm training, and
Current most widely used artificial neural network.BP artificial neural network since 1986 are born, manage by network performance and application
By etc. it is very mature, can satisfy the functional requirement of noise reduction algorithm.The BP artificial neural network that the present invention uses point
For input layer, hidden layer, output layer three-decker, every layer includes multiple neurodes.The present invention uses hearing aid using user
When the voice signal that receives, identify and store it is therein have words section, be supplied to neural network as training data and be trained.
Training data will be constantly updated with the time, to match the variation of user's living environment.
The present invention is divided into hearing aid, smart machine, cloud three parts.Wherein noise reduction algorithm is located at the DSP at hearing aid end
In (Digital Signal Processing, Digital Signal Processing) chip or the dedicated speech processing chip of hearing aid, with
Reach real-time necessary to hearing aid, and can satisfy the demand that off line uses.Neural network algorithm is located at the intelligence such as mobile phone, plate
In the mating APP of hearing aid of energy equipment end, operation is carried out using smart machine hardware, and pass through bluetooth module with hearing aid and carry out
Connection.Hearing aid sends smart machine by bluetooth module for the voice data received, and is stored by smart machine
With processing.Smart machine is trained artificial neural network using these voice data, and by the parameter after training via indigo plant
Tooth is sent to hearing aid, is used with adjusting noise reduction algorithm fitting individual subscriber.User can be with unrestricted choice by voice, nerve net
The data such as network parameter upload to cloud or download from cloud, to realize the data sharing between multiple smart machines.
Present invention be described in more detail with specific example with reference to the accompanying drawing.
Fig. 1 is the topological structure of BP artificial neural network used in the present invention.The major part of the BP neural network is defeated
Enter layer, hidden layer and output layer three-decker.Since the voice signal of input needs to extract voice by mel cepstrum coefficients method
Feature, and voice signal becomes 24 dimensional feature signals after passing through Meier filter group, therefore input layer includes 24 neurodes.
Initial hidden layer includes 25 neurodes, and number of nodes can be adjusted with the variation of training data.Output layer is then
Exported comprising 30 neurodes, and for multichannel, in corresponding noise reduction algorithm under different frequency cross subtracting coefficient, gain compensation because
The parameters such as son.The connection weight of input layer and hidden layer, hidden layer and output layer is respectively ωij,ωjk, hidden layer threshold value a,
Output layer threshold value is b.The neural network that the initial value of these parameters is practiced from training on trial is instructed using a large amount of random voice data
The neural network got out.Training process commonly trains formula using BP artificial neural network, input layer in formula, hidden layer,
Output layer number of nodes is denoted as m, l, n:
Step 1: hidden layer, which is calculated, using general hidden layer excitation function exports, H in formulajWith ajRespectively correspond the hidden of j-th of node
Containing layer output and hidden layer threshold value.
Step 2: calculating output layer error, O in formulakWith ekRespectively correspond the output and output error of k-th of node, YkFor
Desired output from training data.
ek=Yk-Ok
(k=1,2 ..., l)
Step 3: updating weight and threshold value.η is learning rate in formula, can freely be adjusted, bkFor the output of k-th of node
Layer threshold value.
ωjk=ωjk+ηHjek
bk=bk+ek
Continuous iteration is carried out using above-mentioned formula, the number of iterations until reaching setting.
Fig. 2 is the training flow chart of neural network used in the present invention.When user has accumulated foot in hearing aid use process
After enough voice data, neural network algorithm can start to run in smart machine end.BP can be independently set when operation first
The structure of artificial neural network, the number of nodes including the number of plies Yu each layer, and be program by parameter settings such as the threshold value of each layer, weights
Given initial value.Connect it is lower by data by Meier filter group and start train artificial neural network.Scheming this part
It is specifically described in one.After reaching the number of iterations of setting, program can send output parameter to hearing aid and thus
Adjust noise reduction algorithm.Hereafter hearing aid enters the routine use stage, after having accumulated enough new speech data, algorithm meeting
Selection new data and legacy data input neural network according to a certain percentage, are trained again.
Fig. 3 is hearing aid in the present invention, smart machine and cloud relational graph.Hearing aid and smart machine are carried out by bluetooth
Connection, hearing aid can be handled the voice data transmission received to smart machine after connection, and smart machine then may be used
The obtained parameter after neural metwork training is transferred to hearing aid, to modify the noise reduction algorithm configuration of hearing aid.Intelligence is set
Standby then be connected with cloud server by internet, user can upload and download personal data with unrestricted choice, to reach multiple
The shared purpose of device data between intelligence.Also, user can download new firmware packet from cloud server, then be upgraded by bluetooth
The firmware of hearing aid makes hearing aid keep latest function.
Claims (4)
1. a kind of Noise reduction method based on personal data and artificial neural network, characterized in that by analyzing hearing aid
The voice signal that user receives identifies and stores words section, artificial to BP using the sufficient amount of voice messaging of storage
Neural network is trained.By training to the weight of each layer of neural network and optimizing and revising for threshold value, to obtain mature BP
Artificial neural network to correct conventional voice de-noising formula, and then enables hearing aid user obtain optimizing for its people
The noise reduction being more clear after voice.
2. the Noise reduction method based on personal data and artificial neural network as described in claim 1, characterized in that into
One step, noise reduction algorithm is set to the operation of hearing aid end, neural network algorithm is set to smart machine operation, smart machine
It is wirelessly connect with hearing aid.User can independently set voice accumulation, whenever the speech volume of accumulation reaches user
When setting value, hearing aid sends smart machine by bluetooth module for these voice data automatically, and is carried out by smart machine
Storage and processing, smart machine are trained artificial neural network using these voice data, and the parameter after training is sent out
Hearing aid is given, is used with adjusting noise reduction algorithm fitting individual subscriber.User can be joined voice, neural network with unrestricted choice
Number data upload to cloud or download from cloud, to realize the data sharing between multiple smart machines.
3. the Noise reduction method based on personal data and artificial neural network as described in claim 1, characterized in that BP
The topological structure of artificial neural network includes: input layer, hidden layer and output layer three-decker, since the voice signal of input needs
Phonetic feature is extracted by mel cepstrum coefficients method, and voice signal is by becoming 24 dimensional features letter after Meier filter group
Number, therefore input layer includes 24 neurodes, initial hidden layer includes 25 neurodes, and number of nodes can be with
The variation of training data and adjust, output layer then includes 30 neurodes, and is exported for multichannel, in corresponding noise reduction algorithm not
Subtracting coefficient, gain compensation factor parameter, the connection weight point of input layer and hidden layer, hidden layer and output layer are crossed under same frequency
It Wei not ωij,ωjk, hidden layer threshold value a, output layer threshold value is b, ωij、ωjk, a, b these parameters initial value come from training on trial
Experienced neural network, i.e., the neural network obtained using a large amount of random voice data training, training process are manually refreshing using BP
Formula is commonly trained through network, input layer, hidden layer, output layer number of nodes are denoted as m, l, n in formula:
Step 1: hidden layer, which is calculated, using general hidden layer excitation function exports, H in formulajWith ajRespectively correspond the hidden layer of j-th of node
Output and hidden layer threshold value
Step 2: calculating output layer error, O in formulakWith ekRespectively correspond the output and output error of k-th of node, YkFor from
The desired output of training data
ek=Yk-Ok
K=1,2 ..., l
Step 3: updating weight and threshold value, η is learning rate, b in formulakFor the output layer threshold value of k-th of node
ωjk=ωjk+ηHjek
bk=bk+ek
Continuous iteration is carried out using above-mentioned formula, the number of iterations until reaching setting.
4. the Noise reduction method based on personal data and artificial neural network as described in claim 1, characterized in that institute
It is as follows with the training process of neural network: after user has accumulated enough voice data in hearing aid use process, mind
It can start to run in smart machine end through network algorithm, when operation, can independently set the structure of BP artificial neural network first,
Number of nodes including the number of plies Yu each layer, and be the initial value that program gives by parameter settings such as the threshold value of each layer, weights, connect lower incite somebody to action
Data pass through Meier filter group and start to train artificial neural network, and after reaching the number of iterations of setting, program can will be defeated
Parameter is sent to hearing aid and thus adjusts noise reduction algorithm out, and hereafter hearing aid enters the routine use stage, until having accumulated foot
After enough new speech data, algorithm can select new data and legacy data to input neural network according to a certain percentage, carry out again
Training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811602590.XA CN109714692A (en) | 2018-12-26 | 2018-12-26 | Noise reduction method based on personal data and artificial neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811602590.XA CN109714692A (en) | 2018-12-26 | 2018-12-26 | Noise reduction method based on personal data and artificial neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109714692A true CN109714692A (en) | 2019-05-03 |
Family
ID=66258524
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811602590.XA Pending CN109714692A (en) | 2018-12-26 | 2018-12-26 | Noise reduction method based on personal data and artificial neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109714692A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111150422A (en) * | 2020-03-09 | 2020-05-15 | 国微集团(深圳)有限公司 | Intelligent auscultation system and auscultation method thereof |
| CN112562716A (en) * | 2020-12-03 | 2021-03-26 | 兰州交通大学 | Voice enhancement method, device, terminal and medium based on neural network |
| CN112954570A (en) * | 2021-02-20 | 2021-06-11 | 深圳市智听科技有限公司 | Hearing assistance method, device, equipment and medium integrating edge computing and cloud computing |
| CN113812173A (en) * | 2019-05-09 | 2021-12-17 | 索诺瓦有限公司 | Hearing device system and method for processing audio signals |
| CN114765722A (en) * | 2021-01-04 | 2022-07-19 | 大北欧听力公司 | Improving usability and satisfaction of hearing aids |
| CN115116446A (en) * | 2022-06-21 | 2022-09-27 | 成都理工大学 | Method for constructing speaker recognition model in noise environment |
| CN117676445A (en) * | 2024-01-31 | 2024-03-08 | 深圳市鑫正宇科技有限公司 | Hearing aid control method, storage medium and hardware architecture |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1879449A (en) * | 2003-11-24 | 2006-12-13 | 唯听助听器公司 | Hearing aid and a method of noise reduction |
| CN102231281A (en) * | 2011-07-18 | 2011-11-02 | 渤海大学 | Voice visualization method based on integration characteristic and neural network |
| CN105611477A (en) * | 2015-12-27 | 2016-05-25 | 北京工业大学 | Depth and breadth neural network combined speech enhancement algorithm of digital hearing aid |
| CN106714062A (en) * | 2016-11-30 | 2017-05-24 | 天津大学 | BP-artificial-neural-network-based intelligent matching algorithm for digital hearing aid |
| CN106782497A (en) * | 2016-11-30 | 2017-05-31 | 天津大学 | A kind of intelligent sound noise reduction algorithm based on Portable intelligent terminal |
-
2018
- 2018-12-26 CN CN201811602590.XA patent/CN109714692A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1879449A (en) * | 2003-11-24 | 2006-12-13 | 唯听助听器公司 | Hearing aid and a method of noise reduction |
| CN102231281A (en) * | 2011-07-18 | 2011-11-02 | 渤海大学 | Voice visualization method based on integration characteristic and neural network |
| CN105611477A (en) * | 2015-12-27 | 2016-05-25 | 北京工业大学 | Depth and breadth neural network combined speech enhancement algorithm of digital hearing aid |
| CN106714062A (en) * | 2016-11-30 | 2017-05-24 | 天津大学 | BP-artificial-neural-network-based intelligent matching algorithm for digital hearing aid |
| CN106782497A (en) * | 2016-11-30 | 2017-05-31 | 天津大学 | A kind of intelligent sound noise reduction algorithm based on Portable intelligent terminal |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113812173A (en) * | 2019-05-09 | 2021-12-17 | 索诺瓦有限公司 | Hearing device system and method for processing audio signals |
| CN111150422A (en) * | 2020-03-09 | 2020-05-15 | 国微集团(深圳)有限公司 | Intelligent auscultation system and auscultation method thereof |
| CN112562716A (en) * | 2020-12-03 | 2021-03-26 | 兰州交通大学 | Voice enhancement method, device, terminal and medium based on neural network |
| CN114765722A (en) * | 2021-01-04 | 2022-07-19 | 大北欧听力公司 | Improving usability and satisfaction of hearing aids |
| CN112954570A (en) * | 2021-02-20 | 2021-06-11 | 深圳市智听科技有限公司 | Hearing assistance method, device, equipment and medium integrating edge computing and cloud computing |
| CN112954570B (en) * | 2021-02-20 | 2022-10-25 | 深圳市智听科技有限公司 | Hearing assistance method, device, equipment and medium integrating edge computing and cloud computing |
| CN115116446A (en) * | 2022-06-21 | 2022-09-27 | 成都理工大学 | Method for constructing speaker recognition model in noise environment |
| CN115116446B (en) * | 2022-06-21 | 2024-11-01 | 成都理工大学 | Speaker recognition model construction method in noise environment |
| CN117676445A (en) * | 2024-01-31 | 2024-03-08 | 深圳市鑫正宇科技有限公司 | Hearing aid control method, storage medium and hardware architecture |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109714692A (en) | Noise reduction method based on personal data and artificial neural network | |
| CN109326302B (en) | Voice enhancement method based on voiceprint comparison and generation of confrontation network | |
| CN109151692B (en) | Hearing aid self-fitting method based on deep learning network | |
| CN109246515B (en) | A kind of intelligent earphone and method promoting personalized sound quality function | |
| CN106782497B (en) | Intelligent voice noise reduction algorithm based on portable intelligent terminal | |
| CN110428849B (en) | Voice enhancement method based on generation countermeasure network | |
| CN106714062B (en) | An intelligent fitting method for digital hearing aids based on BP artificial neural network | |
| CN106462384A (en) | Multi-modal based intelligent robot interaction method and intelligent robot | |
| WO2020098256A1 (en) | Speech enhancement method based on fully convolutional neural network, device, and storage medium | |
| DE102019206743A1 (en) | Hearing aid system and method for processing audio signals | |
| CN103903627A (en) | Voice-data transmission method and device | |
| DE112020002531T5 (en) | EMOTION DETECTION USING SPEAKER BASELINE | |
| CN107682561A (en) | volume adjusting method, device, terminal and storage medium | |
| CN112653980B (en) | Interactive self-checking and matching method for intelligent hearing aid | |
| CN109671446A (en) | A kind of deep learning sound enhancement method based on absolute hearing threshold | |
| CN105590625A (en) | Acoustic model self-adaptive method and system | |
| CN109859768A (en) | Artificial cochlea's sound enhancement method | |
| CN109831732A (en) | Intelligent chauvent's criterion device and method based on smart phone | |
| CN109448702A (en) | Artificial cochlea's auditory scene recognition methods | |
| CN105791913A (en) | Multi-user telephone configuration information setting method, multi-user telephone configuration information setting system and cloud server | |
| CN109448755A (en) | Artificial cochlea's auditory scene recognition methods | |
| CN113838473A (en) | Voice processing method and device of equipment and equipment | |
| WO2024072700A1 (en) | Switchable noise reduction profiles | |
| KR102655933B1 (en) | Method and apparatus for data augmentation | |
| CN107154258A (en) | Method for recognizing sound-groove based on negatively correlated incremental learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190503 |
|
| RJ01 | Rejection of invention patent application after publication |