CN109448755A - Artificial cochlea's auditory scene recognition methods - Google Patents
Artificial cochlea's auditory scene recognition methods Download PDFInfo
- Publication number
- CN109448755A CN109448755A CN201811276573.1A CN201811276573A CN109448755A CN 109448755 A CN109448755 A CN 109448755A CN 201811276573 A CN201811276573 A CN 201811276573A CN 109448755 A CN109448755 A CN 109448755A
- Authority
- CN
- China
- Prior art keywords
- scene
- ubm
- parameter
- recognition methods
- auditory
- 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 38
- 210000003477 cochlea Anatomy 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 19
- 230000008569 process Effects 0.000 claims abstract description 9
- 238000000605 extraction Methods 0.000 claims abstract description 6
- 238000009432 framing Methods 0.000 claims abstract description 4
- 230000003044 adaptive effect Effects 0.000 claims description 17
- 238000001228 spectrum Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 230000006978 adaptation Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 8
- 238000012545 processing Methods 0.000 abstract description 7
- 230000002708 enhancing effect Effects 0.000 abstract description 2
- 238000005457 optimization Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000007943 implant Substances 0.000 description 2
- 206010011891 Deafness neurosensory Diseases 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 208000009966 Sensorineural Hearing Loss Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000002146 bilateral effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 210000000860 cochlear nerve Anatomy 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000002513 implantation Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 231100000879 sensorineural hearing loss Toxicity 0.000 description 1
- 208000023573 sensorineural hearing loss disease Diseases 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
- 230000004936 stimulating effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; 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
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; 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
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; 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
- G10L25/45—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of analysis window
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses a kind of artificial cochlea's auditory scene recognition methods comprising following steps: (A) establishes the scene training UBM of standard；(B) voice signal is subjected to framing and windowing process；(C) pretreated voice signal is identified by frame；(D) VAD treated scene noise signal is subjected to characteristic vector pickup；(E) signal after feature extraction is handled in GMM-UBM system, obtains likelihood score value, finally identifies scene type.Artificial cochlea's auditory scene recognition methods is by establishing a series of models, it can identify different auditory scenes, instruction is provided for signal processing modules such as the enhancing of speech processor subsequent voice and speech strategies, match the signal processing of speech processor more with auditory scene, improve clarity, the intelligibility of the voice signal of patient in a noisy environment, the listening effect under music scenario also can be improved simultaneously, further improve the quality of life of artificial cave patient.
Description
Technical field
The present invention relates to a kind of auditory scene recognition methods more particularly to a kind of artificial cochlea's auditory scene recognition methods.
Background technique
Artificial cochlea is recognized in the world bilateral severe or pole profound sensorineural hearing loss patient to be made to restore to listen
The unique effective ways and device felt.Existing artificial cochlea's operation workflow are as follows: sound is first converted to telecommunications by microphone acquisition
Number, it by special digitized processing, is encoded according still further to certain strategy, is transmitted to body by being loaded in the transmitting coil after ear
It is interior, it after the receiving coil of implant senses signal, is decoded by decoding chip, the stimulating electrode of implant is made to generate electric current,
To stimulate auditory nerve to generate the sense of hearing.Due to the limitation of use environment, environment noise is necessarily adulterated in sound, is needed to sound
Signal carries out certain algorithm optimization, but in view of the diversification of use environment, if only using single algorithm optimization, algorithm is excellent
Signal after change is deviated with actual conditions sometimes, is unable to reach optimal auditory effect, therefore needs a kind of auditory scene
Recognition methods so that different scenes use different optimization algorithms, have reached optimal auditory effect.
Summary of the invention
In view of the above drawbacks of the prior art, technical problem to be solved by the invention is to provide a kind of artificial cochleas to listen
Feel scene recognition method, can identify different auditory scenes.
To achieve the above object, the present invention provides a kind of artificial cochlea's auditory scene recognition methods comprising following step
Rapid: the various scene training signals of (A) model training program module collection form the scene training UBM of standard by EM algorithm；
(B) voice signal is carried out framing and windowing process by preprocessor module；(C) VAD handler module is to pretreated
Voice signal is identified by frame, identifies that the frame signal is scene noise signal or voice signal；(D) feature extraction program mould
VAD treated scene noise signal is carried out characteristic vector pickup by block；(E) scene Recognition program module will be after feature extraction
A part input UBM carries out related operation；A part input GMM operation, then in the related data in UBM and GMM data into
Row operation forms new GMM；The data in UBM are compared with the data in new GMM later, obtain likelihood score value, final to know
It Chu not scene type.
In stepb, which uses Hamming window or Hanning window.
Further, Hamming window:Wherein, the long N of window
=256, frame pipettes 128.
In step C, which uses the VAD detection method based on short-time energy and short-time zero-crossing rate.
In step D, this feature vector, which extracts, uses MFCC or FBank.
Further, it the calculation method of the MFCC parameter of a frame scene noise signal: is calculated and is believed according to discrete Fourier transform
Number discrete spectrum { S (ω) }；Frequency is divided into D=30 equal part by Bark scale, and calculates its centre frequency and edge frequency,
Wherein, Bark scale Ω is with frequency f transformation relationIt is filtered using D triangle band logical
Wave device does logarithmic energy output E (d) (d=1,2 ..., D) that convolution finds out each frequency range with discrete spectrum { S (ω) } respectively,
In, the centre frequency and edge frequency of triangular filter are aligned with corresponding Bark frequency range；Logarithmic energy output to each frequency range
Discrete cosine transform is carried out to obtainIt takes
Preceding 16 dimension is used as characteristic parameter.
In step E, in GMM-UBM system, scene noise model modifies certain of UBM by Bayesian adaptation method
A little parameters obtain, and adaptive algorithm is divided into two steps, and the first step is expectation process, calculate scene training data in each single Gauss of UBM
Statistical parameter in distribution；Second step obtains the parameter of scene noise model with the parameter weighting of new statistical parameter and UBM,
Method of weighting makes in final scene noise model, by adaptive its parameter of distribution of more scene training data close to survey
Try the parameter of scene noise itself, and by its adaptive distribution parameter of less test data close to the parameter of UBM.
Further, UBM and trained vector sequence X={ x is given1,x2,...,xT, each characteristic vector is calculated first to be belonged to
The probability of any Gaussian Profile in UBM calculates i-th of Gaussian Profile in UBMSo
Afterwards according to Pr (i | xt) and xtIt calculates for modifying weight, mean value and the statistical parameter of variance Finally, obtained by scene training data
These new statistical parameters are used to update the model parameter of UBM Auto-adaptive parameter aiControl is new
The balance of old parameter, scale factor γ adjust weight, so that the sum of weight all after adaptive is 1, for i-th of Gauss
Distribution is used for above-mentioned auto-adaptive parameter ai, it is defined asWherein, r is a fixed value, controls the parameter of UBM
Weight in adaptive, sets r=16.
Artificial cochlea's auditory scene recognition methods of the present invention can identify different sense of hearing fields by establishing a series of models
Scape provides instruction for signal processing modules such as the enhancing of speech processor subsequent voice and speech strategies, makes the letter of speech processor
Number processing is more matched with auditory scene, exports the stimulus signal being more consistent with practical auditory scene, patient is in noise for raising
Clarity, the intelligibility of voice signal under environment, while the listening effect under music scenario also can be improved, further improve people
The quality of life of work cochlea implantation patient.
It is described further below with reference to technical effect of the attached drawing to design of the invention, specific structure and generation, with
It is fully understood from the purpose of the present invention, feature and effect.
Detailed description of the invention
Fig. 1 is the flow diagram of artificial cochlea's auditory scene recognition methods of the present invention.
Specific embodiment
The present invention provides a kind of artificial cochlea's auditory scene recognition methods, different auditory scene for identification, such as
Classroom, street, music hall, market, railway station, food market etc..
Artificial cochlea's auditory scene recognition methods includes model training, pretreatment, VAD (Voice Activity
Detection, voice activity detection) processing, feature extraction, five steps of scene Recognition.
Model training: the various scene training signals of model training program module collection (i.e. scene voice signal) establish field
Jing Ku forms the scene training UBM of standard by EM (Expectation Maximization, greatest hope) algorithm
(Universal Background Model, universal background model).
EM algorithm:
Feature vector set o=(o1,o2,...,oT)；
Model λ:
GMM (Gaussian Mixture Model, gauss hybrid models) is distributed maximum likelihood function:
The weights omega of m-th of Gaussm:
The mean value of m-th of Gauss
The variance of m-th of Gauss
Pretreatment: voice signal is carried out framing and windowing process by preprocessor module.
By taking system sampling frequency is 16kHz as an example.
The windowing process uses Hamming window, the long N=256 of window, and frame pipettes the long half of window, i.e., and 128.
Hamming window:
Other window functions such as Hanning window also can be used in the windowing process, and frame length and frame shifting can also be according to system need
It is changed setting.
VAD processing: VAD handler module is identified pretreated voice signal by frame, identifies the frame signal
For scene noise signal or voice signal, wherein the identification uses the detection side VAD based on short-time energy and short-time zero-crossing rate
Method.
Feature extraction: VAD treated scene noise signal is carried out characteristic vector pickup by feature extraction program module,
In, this feature vector, which extracts, uses MFCC (Mel-Frequency Cepstrum Coefficient, mel-frequency cepstrum system
Number) or FBank (Mel-scale Filter Bank, Meier scale filter group).
The calculation method of the MFCC parameter of one frame scene noise signal is as follows:
(1) according to discrete Fourier transform calculate signal discrete spectrum S (ω) | ω=1,2 ..., N }；
(2) frequency is divided into D=30 equal part by Bark scale, and calculates its centre frequency and edge frequency, Bark is carved
Spending Ω with frequency f transformation relation is
(3) the logarithm energy that convolution finds out each frequency range is done with discrete spectrum { S (ω) } respectively using D triangle bandpass filter
Amount output E (d) (d=1,2 ..., D), the wherein centre frequency and edge frequency of triangular filter and corresponding Bark frequency range pair
Together；
(4) discrete cosine transform is carried out to the logarithmic energy output of each frequency range to obtain
Take preceding 16 dimension as characteristic parameter.
Scene Recognition: a part input UBM after feature extraction is carried out related operation by scene Recognition program module；One
Point input GMM operation, then data carry out operation in the related data in UBM and GMM, form new GMM；Later in UBM
Data are compared with the data in new GMM, are obtained likelihood score value, are finally identified scene type, wherein in GMM-UBM system
In system, scene noise model is obtained by certain parameters that Bayesian adaptation method modifies UBM, and adaptive algorithm is divided into two
Step, the first step is expectation process, calculates statistical parameter of the scene training data in each single Gaussian Profile of UBM；Second step, with new
Statistical parameter and the parameter weighting of UBM obtain the parameter of scene noise model, method of weighting makes final scene noise mould
In type, by more scene training data it is adaptive be distributed its parameter close to test scene noise itself parameter, and by compared with
Parameter of few its adaptive distribution parameter of test data close to UBM.
Adaptive approach is as follows, gives UBM and trained vector sequence X={ x1,x2,...,xT, each feature is calculated first
Vector belongs to the probability of any Gaussian Profile in UBM.To i-th of Gaussian Profile in UBM, calculate
Then according to Pr (i | xt) and xtIt calculates for modifying weight, mean value and the statistical parameter of variance
Finally, these the new statistical parameters obtained by scene training data are used to update the model parameter of UBM
Auto-adaptive parameter aiThe balance of new and old parameter is controlled, scale factor γ adjusts weight, so that all after adaptive
The sum of weight be 1.
For i-th of Gaussian Profile, it to be used for above-mentioned auto-adaptive parameter ai, it is defined as
Wherein, r is a fixed value, controls the weight of the parameter of UBM in adaptive, sets r=16.Using data phase
The auto-adaptive parameter of pass, so that being adaptively related to Gaussian Profile.If the probability number n of a distributioniIt is smaller, then ai
→ 0, parameter of the adaptive rear scene noise model parameters close to UBM.If the probability number n of a distributioniIt is bigger, then
ai→ 1, scene noise model parameter is mainly determined by scene training data.
The preferred embodiment of the present invention has been described in detail above.It should be appreciated that those skilled in the art without
It needs creative work according to the present invention can conceive and makes many modifications and variations.Therefore, all technologies in the art
Personnel are available by logical analysis, reasoning, or a limited experiment on the basis of existing technology under this invention's idea
Technical solution, all should be within the scope of protection determined by the claims.
Claims (8)
1. a kind of artificial cochlea's auditory scene recognition methods comprising following steps: (A) model training program module collection is various
Scene training signal forms the scene training UBM of standard by EM algorithm；(B) preprocessor module carries out voice signal
Framing and windowing process；(C) VAD handler module is identified pretreated voice signal by frame, identifies that the frame is believed
Number be scene noise signal or voice signal；(D) feature extraction program module carries out VAD treated scene noise signal
Characteristic vector pickup；(E) a part input UBM after feature extraction is carried out related operation by scene Recognition program module；One
Point input GMM operation, then data carry out operation in the related data in UBM and GMM, form new GMM；Later in UBM
Data are compared with the data in new GMM, are obtained likelihood score value, are finally identified scene type.
2. artificial cochlea's auditory scene recognition methods as described in claim 1, it is characterised in that: in stepb, at the adding window
Reason uses Hamming window or Hanning window.
3. artificial cochlea's auditory scene recognition methods as claimed in claim 2, it is characterised in that: Hamming window:Wherein, the long N=256 of window, frame pipette 128.
4. artificial cochlea's auditory scene recognition methods as described in claim 1, it is characterised in that: in step C, which is adopted
With the VAD detection method based on short-time energy and short-time zero-crossing rate.
5. artificial cochlea's auditory scene recognition methods as described in claim 1, it is characterised in that: in step D, this feature to
Amount, which is extracted, uses MFCC or FBank.
6. artificial cochlea's auditory scene recognition methods as claimed in claim 5, it is characterised in that: a frame scene noise signal
The calculation method of MFCC parameter: the discrete spectrum { S (ω) } of signal is calculated according to discrete Fourier transform；By Bark scale frequency
It is divided into D=30 equal part, and calculates its centre frequency and edge frequency, wherein Bark scale Ω is with frequency f transformation relationConvolution is done with discrete spectrum { S (ω) } respectively using D triangle bandpass filter to find out
The logarithmic energy of each frequency range exports E (d) (d=1,2 ..., D), wherein the centre frequency and edge frequency of triangular filter
It is aligned with corresponding Bark frequency range；Discrete cosine transform is carried out to the logarithmic energy output of each frequency range to obtainTake preceding 16 dimension as characteristic parameter.
7. artificial cochlea's auditory scene recognition methods as described in claim 1, it is characterised in that: in step E, in GMM-
In UBM system, scene noise model is obtained by certain parameters that Bayesian adaptation method modifies UBM, adaptive algorithm point
For two steps, the first step is expectation process, calculates statistical parameter of the scene training data in each single Gaussian Profile of UBM；Second step,
The parameter of scene noise model is obtained with the parameter weighting of new statistical parameter and UBM, method of weighting makes final scene make an uproar
In acoustic model, by adaptive its parameter of distribution of more scene training data close to the parameter of test scene noise itself, and
By its adaptive distribution parameter of less test data close to the parameter of UBM.
8. artificial cochlea's auditory scene recognition methods as claimed in claim 7, it is characterised in that: given UBM and trained vector
Sequence X={ x1,x2,...,xT, the probability that each characteristic vector belongs to any Gaussian Profile in UBM is calculated first, in UBM
I-th of Gaussian Profile calculatesThen according to Pr (i | xt) and xtIt calculates and is used for the power of amendment
The statistical parameter of weight, mean value and variance Finally, these the new statistical parameters obtained by scene training data are used to update
The model parameter of UBM Auto-adaptive parameter aiControl the balance of new and old parameter, scale factor γ adjustment
Weight, for i-th of Gaussian Profile, is used for above-mentioned auto-adaptive parameter a so that the sum of weight all after adaptive is 1i,
It is defined asWherein, r is a fixed value, controls the weight of the parameter of UBM in adaptive, sets r=16.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811276573.1A CN109448755A (en) | 2018-10-30 | 2018-10-30 | Artificial cochlea's auditory scene recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811276573.1A CN109448755A (en) | 2018-10-30 | 2018-10-30 | Artificial cochlea's auditory scene recognition methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109448755A true CN109448755A (en) | 2019-03-08 |
Family
ID=65548788
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811276573.1A Pending CN109448755A (en) | 2018-10-30 | 2018-10-30 | Artificial cochlea's auditory scene recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109448755A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109859768A (en) * | 2019-03-12 | 2019-06-07 | 上海力声特医学科技有限公司 | Artificial cochlea's sound enhancement method |
| CN109893340A (en) * | 2019-03-25 | 2019-06-18 | 深圳信息职业技术学院 | A kind of processing method and processing device of the voice signal of cochlear implant |
| CN109979477A (en) * | 2019-03-12 | 2019-07-05 | 上海力声特医学科技有限公司 | The sound processing method of artificial cochlea |
| CN112820318A (en) * | 2020-12-31 | 2021-05-18 | 西安合谱声学科技有限公司 | Impact sound model establishment and impact sound detection method and system based on GMM-UBM |
| CN113038344A (en) * | 2019-12-09 | 2021-06-25 | 三星电子株式会社 | Electronic device and control method thereof |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101241699A (en) * | 2008-03-14 | 2008-08-13 | 北京交通大学 | A speaker identification system for remote Chinese teaching |
| CN106251861A (en) * | 2016-08-05 | 2016-12-21 | 重庆大学 | A kind of abnormal sound in public places detection method based on scene modeling |
| CN106941005A (en) * | 2017-02-24 | 2017-07-11 | 华南理工大学 | A kind of vocal cords method for detecting abnormality based on speech acoustics feature |
| CN106952643A (en) * | 2017-02-24 | 2017-07-14 | 华南理工大学 | A kind of sound pick-up outfit clustering method based on Gaussian mean super vector and spectral clustering |
| CN107103901A (en) * | 2017-04-03 | 2017-08-29 | 浙江诺尔康神经电子科技股份有限公司 | Artificial cochlea's sound scenery identifying system and method |
| DE102016214745A1 (en) * | 2016-08-09 | 2018-02-15 | Carl Von Ossietzky Universität Oldenburg | Method for stimulating an implanted electrode arrangement of a hearing prosthesis |
| CN108231067A (en) * | 2018-01-13 | 2018-06-29 | 福州大学 | Sound scenery recognition methods based on convolutional neural networks and random forest classification |
| CN108305616A (en) * | 2018-01-16 | 2018-07-20 | 国家计算机网络与信息安全管理中心 | A kind of audio scene recognition method and device based on long feature extraction in short-term |
-
2018
- 2018-10-30 CN CN201811276573.1A patent/CN109448755A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101241699A (en) * | 2008-03-14 | 2008-08-13 | 北京交通大学 | A speaker identification system for remote Chinese teaching |
| CN106251861A (en) * | 2016-08-05 | 2016-12-21 | 重庆大学 | A kind of abnormal sound in public places detection method based on scene modeling |
| DE102016214745A1 (en) * | 2016-08-09 | 2018-02-15 | Carl Von Ossietzky Universität Oldenburg | Method for stimulating an implanted electrode arrangement of a hearing prosthesis |
| CN106941005A (en) * | 2017-02-24 | 2017-07-11 | 华南理工大学 | A kind of vocal cords method for detecting abnormality based on speech acoustics feature |
| CN106952643A (en) * | 2017-02-24 | 2017-07-14 | 华南理工大学 | A kind of sound pick-up outfit clustering method based on Gaussian mean super vector and spectral clustering |
| CN107103901A (en) * | 2017-04-03 | 2017-08-29 | 浙江诺尔康神经电子科技股份有限公司 | Artificial cochlea's sound scenery identifying system and method |
| CN108231067A (en) * | 2018-01-13 | 2018-06-29 | 福州大学 | Sound scenery recognition methods based on convolutional neural networks and random forest classification |
| CN108305616A (en) * | 2018-01-16 | 2018-07-20 | 国家计算机网络与信息安全管理中心 | A kind of audio scene recognition method and device based on long feature extraction in short-term |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109859768A (en) * | 2019-03-12 | 2019-06-07 | 上海力声特医学科技有限公司 | Artificial cochlea's sound enhancement method |
| CN109979477A (en) * | 2019-03-12 | 2019-07-05 | 上海力声特医学科技有限公司 | The sound processing method of artificial cochlea |
| CN109893340A (en) * | 2019-03-25 | 2019-06-18 | 深圳信息职业技术学院 | A kind of processing method and processing device of the voice signal of cochlear implant |
| CN113038344A (en) * | 2019-12-09 | 2021-06-25 | 三星电子株式会社 | Electronic device and control method thereof |
| CN112820318A (en) * | 2020-12-31 | 2021-05-18 | 西安合谱声学科技有限公司 | Impact sound model establishment and impact sound detection method and system based on GMM-UBM |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109448755A (en) | Artificial cochlea's auditory scene recognition methods | |
| CN112509564B (en) | End-to-end voice recognition method based on connection time sequence classification and self-attention mechanism | |
| WO2019232829A1 (en) | Voiceprint recognition method and apparatus, computer device and storage medium | |
| CN108447495B (en) | Deep learning voice enhancement method based on comprehensive feature set | |
| US8842853B2 (en) | Pitch perception in an auditory prosthesis | |
| Stern et al. | Hearing is believing: Biologically inspired methods for robust automatic speech recognition | |
| CN106782565A (en) | A kind of vocal print feature recognition methods and system | |
| CN110428842A (en) | Speech model training method, device, equipment and computer readable storage medium | |
| JP2022529641A (en) | Speech processing methods, devices, electronic devices and computer programs | |
| CN109328380B (en) | Recursive noise power estimation with noise model adaptation | |
| CN105513605A (en) | Voice enhancement system and method for cellphone microphone | |
| WO2020087716A1 (en) | Auditory scene recognition method for artificial cochlea | |
| CN109121057A (en) | A kind of method and its system of intelligence hearing aid | |
| CN102509547A (en) | Method and system for voiceprint recognition based on vector quantization based | |
| CN108922541A (en) | Multidimensional characteristic parameter method for recognizing sound-groove based on DTW and GMM model | |
| CN110111769B (en) | Electronic cochlea control method and device, readable storage medium and electronic cochlea | |
| CN109859768A (en) | Artificial cochlea's sound enhancement method | |
| CN104778948B (en) | A kind of anti-noise audio recognition method based on bending cepstrum feature | |
| CN112151056A (en) | Intelligent cochlear sound processing system and method with customization | |
| CN109243466A (en) | A kind of vocal print authentication training method and system | |
| ES2849124A1 (en) | Artificial cochlea ambient sound sensing method and system | |
| CN112017658A (en) | Operation control system based on intelligent human-computer interaction | |
| CN111489763A (en) | Adaptive method for speaker recognition in complex environment based on GMM model | |
| Gandhiraj et al. | Auditory-based wavelet packet filterbank for speech recognition using neural network | |
| Zezario et al. | Speech enhancement with zero-shot model selection |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190308 |
|
| WD01 | Invention patent application deemed withdrawn after publication |