CN108182950B - Improved method for decomposing and extracting abnormal sound characteristics of public places through empirical wavelet transform - Google Patents
Improved method for decomposing and extracting abnormal sound characteristics of public places through empirical wavelet transform Download PDFInfo
- Publication number
- CN108182950B CN108182950B CN201711462639.1A CN201711462639A CN108182950B CN 108182950 B CN108182950 B CN 108182950B CN 201711462639 A CN201711462639 A CN 201711462639A CN 108182950 B CN108182950 B CN 108182950B
- Authority
- CN
- China
- Prior art keywords
- abnormal sound
- sound signal
- wavelet transform
- empirical wavelet
- original
- 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.)
- Active
Links
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 78
- 238000000034 method Methods 0.000 title claims abstract description 21
- 230000005236 sound signal Effects 0.000 claims abstract description 32
- 239000013598 vector Substances 0.000 claims abstract description 16
- 238000001228 spectrum Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000010606 normalization Methods 0.000 claims abstract description 6
- 238000000605 extraction Methods 0.000 claims description 11
- 238000000354 decomposition reaction Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000000844 transformation Methods 0.000 claims 1
- 230000009466 transformation Effects 0.000 claims 1
- 238000012544 monitoring process Methods 0.000 description 12
- 230000006870 function Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 206010039740 Screaming Diseases 0.000 description 3
- 238000013145 classification model Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 206010011224 Cough Diseases 0.000 description 1
- 210000000721 basilar membrane Anatomy 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
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
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/18—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
-
- 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
- G10L15/00—Speech recognition
- G10L15/02—Feature extraction for speech recognition; Selection of recognition unit
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Soundproofing, Sound Blocking, And Sound Damping (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention provides an improved method for decomposing and extracting the abnormal sound characteristics of public places by empirical wavelet transform, which is characterized in that the Fourier frequency spectrum of an original abnormal sound signal is divided by a method of simulating equivalent rectangular bandwidth scale, an equivalent rectangular bandwidth empirical wavelet transform filter bank with fixed boundaries is obtained according to the divided boundaries, and the filter bank is used for decomposing the original abnormal sound signal to obtain different frequency components of the original abnormal sound signal; and calculating the energy entropy of different frequency components of the original abnormal sound signal relative to the original abnormal sound signal, and taking the energy entropy after normalization processing as the feature vector of the original abnormal sound signal.
Description
Technical Field
The invention belongs to the technical field of audio signal feature extraction and identification.
Background
Public places such as squares, bus stations, subways and the like have the characteristics of dense crowd gathering, wide regions and the like, and the safety precaution of the public places is always concerned by governments and people of all countries. At present, a monitoring technology mainly based on video monitoring plays an active role in safety precaution in public places, but the video monitoring technology has the problems of monitoring dead angles, monitoring fuzziness in rainy days and the like. As is well known, abnormal sounds such as screaming sound, gunshot sound, glass breaking sound, explosion sound and the like are often accompanied when an abnormal event occurs, and therefore the cooperative operation of audio monitoring and video monitoring has become a development direction in the field of security monitoring in public places. At present, video monitoring is mature relative to audio monitoring and is widely applied to practice, and the existing audio monitoring system only comprises simple sound collection, transmission and the like and lacks effective identification of abnormal sounds. The technology for recognizing the abnormal sound in the public places is a core technology of an audio monitoring system. Therefore, the method has more important social significance and research value for the research of the technology.
The problem of the extraction of abnormal sound characteristics in public places is solved by Empirical Wavelet Transform (EWT): empirical wavelet transform is performed by detecting the minimum between adjacent maxima of the fourier spectrum of the signal as a boundary and constructing a set of filter banks with such a set of boundaries, as shown in fig. 1. The method for detecting the Fourier spectrum extreme point of the signal is suitable for the signal with simple frequency components, and after certain noise is added to the signal, the detection result is greatly influenced. And the abnormal sound of the public place is composed of two parts: abnormal sound signals and background noise signals. The Fourier spectrum of the abnormal sound signal contains more extreme points, and due to the existence of background noise, when the EWT is adopted to decompose the abnormal sound signal in the public place, the extreme points of the Fourier spectrum of the signal are complex, so that the distribution of the constructed filter bank is relatively concentrated, and the distribution of the constructed filter bank is also concentrated. When this method is used to identify abnormal sounds in public places, undesirable results are obtained. Secondly, due to the fact that background noise of public places is complex, the obtained results of signal decomposition of the same type have large difference, and the same result of classification which is not ideal can be obtained. Meanwhile, due to the fact that the signal is easily interfered by noise, reconstruction errors of the signal are large.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a public place abnormal sound feature decomposition and extraction based on improved Empirical Wavelet Transform, the public place abnormal sound is decomposed by improved Empirical Wavelet Transform (DEWT), the key theory and the technical problem that the EWT is applied to the public place abnormal sound feature extraction are solved by introducing Equivalent Rectangular Bandwidth (ERB) in the process, and the features of the public place abnormal sound under different scales are obtained.
The invention provides an improved public place abnormal sound feature decomposition and extraction method based on empirical wavelet transform, which comprises the following steps:
And 2, calculating the energy entropy of different frequency components of the original abnormal sound signal relative to the original abnormal sound signal, and taking the energy entropy after normalization processing as the feature vector of the original abnormal sound signal.
The specific method of the step 1 is as follows:
(1) calculating the center frequency f of an equivalent rectangular bandwidth filtercLet the center frequency f of the filter in the empirical wavelet transformcnEqual to the center frequency f of the equivalent rectangular bandwidth filterc。
(2) Calculation formula f of upper and lower boundaries of filter in empirical wavelet transformcn＝(wn+wn+1) /2, obtaining a series of boundary values wnAt a boundary value of wnAnd substituting the obtained signal into a low-pass filter and a band-pass filter bank constructed by empirical wavelets to form an equivalent rectangular bandwidth empirical wavelet transform filter bank.
(3) Decomposing each original abnormal sound signal through the equivalent rectangular bandwidth empirical wavelet transform filter bank to obtain different frequency components f of the original abnormal sound signalj。
The specific method of the step 2 comprises the following steps:
the improved empirical wavelet transform filter bank is used for decomposing abnormal sounds in public places to obtain a series of different modes fk. Defining an energy entropy HjAs public place abnormal sound feature vector
In the formula, HjRepresenting the entropy of energy, parameter εj＝EjE, E represents the energy of the original abnormal sound signal, Ej＝|fj|2Denotes the jth frequency component fjThe energy of (a). And calculating the energy entropy of different frequency components of the signal to the original abnormal sound signal, and combining the energy entropy into a feature vector to perform normalization processing to serve as the feature vector of the original signal.
The invention has the advantages that:
1. the invention fully considers the key theory and technical problems of applying the EWT method to the identification of the abnormal sound in the public place, simultaneously combines the characteristics of the abnormal sound and the background noise, researches the problem of low identification rate of the EWT to decompose the abnormal sound in the public place, and forms an improved empirical wavelet transform filter bank by simulating the ERB scale.
2. The invention relates to a method more suitable for describing the characteristics of abnormal sound signals in public places.
Drawings
FIG. 1: the signal Fourier spectrum division and the filter bank establishment are shown schematically;
FIG. 2: an ERB-EWT filter bank frequency response graph;
FIG. 3: a scream sound decomposition result graph;
FIG. 4: the invention provides an improved empirical wavelet transform feature extraction flow block diagram;
FIG. 5: the invention provides a flow diagram for recognizing abnormal sounds in public places.
Detailed Description
The whole process of the abnormal sound identification in the public place is taken as an example, and the idea of the invention is further explained by combining the attached drawings:
as shown in fig. 5, the flow chart of the public place abnormal sound identification mainly includes two parts: and carrying out feature extraction on the abnormal sounds in the public places and identifying the abnormal sounds in the public places, wherein the feature extraction is obtained by adopting the improved empirical wavelet transform-based method.
Public place abnormal sound feature extraction
Firstly, simulating ERB scale to obtain filter boundary
Empirical wavelet transform is a method for analyzing nonlinear non-stationary signals, where f is the time domain discrete signal, EWT takes the minimum value between consecutive maxima of the Fourier spectrum of the signal as the boundary of a filter, and the result of the boundary is denoted wnN is an integer, N belongs to [0, N ]]. Assuming that the signal is composed of N modes, the Fourier spectrum range [0, π]The number of boundaries to be determined is N-1, divided into N consecutive segments. w is a0＝0，wnEach segmented segment is denoted as Λ ═ pin＝[wn,wn+1]. Therefore, the first and second electrodes are formed on the substrate,
obtaining the boundary wnThen at wnIs a center 2 taunAn empirical scale function and an empirical wavelet function of the filter are constructed over the region of (a). As shown in FIG. 1, a low pass filter is obtained
And N-1 band-pass filter
Thus forming a bank of filter banks in the frequency domain. A low pass filter and an N-1 band pass filter. However, when the original empirical wavelet transform method is used for processing abnormal sounds, the defects that the recognition rate is low and the decomposition is easily interfered by noise exist. By reading the literature, it was found that the introduction of ERB scale could better solve the problem.
The ERB scale can accurately describe the frequency selective characteristics of the basilar membrane of the human ear and can better reflect the essence of the sound signal. Moore states that the center frequency of the human ear hearing filter covers only 50-15000 Hz. The sampling frequency of the abnormal sound of the public place in the example is 16KHz, and the frequency range is 0-8 KHz.
Center frequency f of human ear auditory filtercThe relation to the scale r of the human auditory filter ERB is:
the center frequency of the human ear auditory filter is calculated as:
the frequency domain range of the abnormal sound is [0,8KHz ]]Then the ERB scale r may be taken to be an integer of 31. Calculating the center frequency f of an equivalent rectangular bandwidth filtercBy making the center frequency of the filter in the empirical wavelet transform and the center frequency f of the equivalent rectangular bandwidth auditory filtercnEqual;
namely: f. ofcn＝fc (3)
Calculating a series of boundary values w by using the center frequency calculated from the upper and lower boundaries of the filter in the empirical wavelet transformn。
fcn＝(wn+wn+1)/2 (4)
From this a series of boundary values w are calculated and obtainednThis value is substituted into the filter bank of the empirical wavelet construction. The first filter in the constructed filter bank is a low-pass filter, and the rest are band-pass filters. Since the start boundary of the EWT-constructed filter bank is 0Hz and the end boundary is 8KHz, the entire frequency range of abnormal sounds is covered.
Wherein the low pass filter is:
the band-pass filter is as follows:
wherein tau isn＝γωnThe function β (x) is an arbitrary Ck([0,1]) A function. Beta (x) ═ x4(35-84x+70x2-20x3). When in use
When the temperature of the water is higher than the set temperature,
at L2(R) spatially constitutes a tight supporting frame.
The approximation coefficient can be obtained by inner product of the signal and an empirical scale function
Detail coefficients can be obtained from the inner product of the signal and an empirical wavelet function
Wherein,
and
are respectively phi1(t) and psin(t) Fourier transform.
For this purpose, the reconstructed signal is obtained by equation (7):
wherein, represents the convolution of the data,
and
are respectively
And
the fourier transform of (d).
Improved empirical wavelet transform filter bank
Forming an improved empirical wavelet transform filter bank by simulating the ERB scale, constructing the filter bank as shown in FIG. 2, wherein the signal to be processed passes through the improved empirical wavelet transform filter bank, and the signal f is decomposed to obtain an empirical mode fkAs shown in formula (8).
Taking screaming sound as an example, the screaming sound is decomposed by using an improved empirical wavelet transform filter group, and the result is shown in fig. 3.
③ extraction of characteristics
The flow of the improved empirical wavelet transform filter bank for extracting the abnormal sound features of the public places is shown in fig. 4, and the improved empirical wavelet transform filter bank is used for decomposing the abnormal sounds of the public places to obtain a series of different modes fk. The energy of the original signal is E, the j-th mode fjCorresponding energy is Ej＝|fj|2. Let epsilonj＝Ej/E, defining the energy entropy HjAs public place abnormal sound feature vector
And calculating the energy entropy of different frequency components of the signal to the original abnormal sound signal, and combining the energy entropy into a feature vector to perform normalization processing to serve as the feature vector of the original signal.
Second, identification of abnormal sound in public place
The identification of abnormal sounds in public places is divided into two steps:
firstly, establishing a classification model: establishing an abnormal sound database, wherein the abnormal sound database comprises five types of similar normal sounds (the similar normal sounds comprise the sound of an automatic teller Machine voice, the sound of a bank call number, the sound of a dog barking, the sound of a drumbeat, the sound of a cough, the sound of opening and closing a door, the sound of knocking and the sound of talking), randomly selecting half of the abnormal sounds from each type as training samples, taking the other half of the abnormal sounds as test samples to be identified, respectively obtaining the feature vectors of the training samples and the test samples by using the method for extracting the abnormal sounds in the public places, and establishing a Support Vector Machine (SVM) classification model according to the feature vectors of the training samples.
Secondly, classification and identification: training an SVM classification model by using the feature vectors obtained by the training samples, testing the samples to be tested according to the model, identifying the types of abnormal sounds of the tested samples, outputting the classification result of each type of abnormal sounds, and obtaining the final identification result.
Claims (2)
1. The improved public place abnormal sound feature decomposition and extraction method based on empirical wavelet transform is characterized by comprising the following steps of:
step 1, dividing a Fourier spectrum of an original abnormal sound signal by a method of simulating equivalent rectangular bandwidth scale, obtaining an equivalent rectangular bandwidth empirical wavelet transform filter bank with a fixed boundary according to the divided boundary, and decomposing the original abnormal sound signal by using the equivalent rectangular bandwidth empirical wavelet transform filter bank to obtain different frequency components of the original abnormal sound signal;
step 2, calculating the energy entropy of different frequency components of the original abnormal sound signal relative to the original abnormal sound signal, and combining the energy entropy into a feature vector to carry out normalization processing to serve as the feature vector of the original abnormal sound signal;
the specific method of the step 1 is as follows:
(1) calculating the center frequency f of an equivalent rectangular bandwidth filtercLet the center frequency of the filter in the empirical wavelet transformfcnEqual to the center frequency f of the equivalent rectangular bandwidth filterc；
(2) Calculation formula f of upper and lower boundaries of filter in empirical wavelet transformcn＝(wn+wn+1) /2, obtaining a series of boundary values wnAt a boundary value of wnSubstituting the obtained signal into a low-pass filter and a band-pass filter bank constructed by empirical wavelets to form an equivalent rectangular bandwidth empirical wavelet transform filter bank;
(3) decomposing each original abnormal sound signal through the equivalent rectangular bandwidth empirical wavelet transform filter bank to obtain different frequency components f of the original abnormal sound signalj。
2. The improved method for decomposing and extracting the abnormal sound features in the public places with small experience transformations according to claim 1, wherein the step 2 specifically comprises the following steps:
(1) the improved empirical wavelet transform filter bank is used for decomposing abnormal sounds in public places to obtain a series of different modes fk；
(2) Defining an energy entropy HjAs public place abnormal sound feature vector
In the formula, HjRepresenting the entropy of energy, parameter εj＝EjE, E represents the energy of the original abnormal sound signal, Ej＝|fj|2Denotes the jth frequency component fjThe energy of (a);
(3) and calculating the energy entropy of different frequency components of the signal to the original abnormal sound signal, and combining the energy entropy into a feature vector to perform normalization processing to serve as the feature vector of the original abnormal sound signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711462639.1A CN108182950B (en) | 2017-12-28 | 2017-12-28 | Improved method for decomposing and extracting abnormal sound characteristics of public places through empirical wavelet transform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711462639.1A CN108182950B (en) | 2017-12-28 | 2017-12-28 | Improved method for decomposing and extracting abnormal sound characteristics of public places through empirical wavelet transform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108182950A CN108182950A (en) | 2018-06-19 |
| CN108182950B true CN108182950B (en) | 2021-05-28 |
Family
ID=62548499
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711462639.1A Active CN108182950B (en) | 2017-12-28 | 2017-12-28 | Improved method for decomposing and extracting abnormal sound characteristics of public places through empirical wavelet transform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108182950B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109034127B (en) * | 2018-08-31 | 2022-04-05 | 中国电子科技集团公司第三十六研究所 | Frequency spectrum anomaly detection method and device and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1207170A (en) * | 1996-09-13 | 1999-02-03 | 株式会社东芝 | Unsteady signal analyzer and medium for recording unsteady signal analyzer program |
| CN103366734A (en) * | 2012-03-31 | 2013-10-23 | 佳能株式会社 | Method and device for checking voice recognition results, voice recognition system and audio monitoring system |
| CN106228979A (en) * | 2016-08-16 | 2016-12-14 | 重庆大学 | A kind of abnormal sound in public places feature extraction and recognition methods |
| CN106683687A (en) * | 2016-12-30 | 2017-05-17 | 杭州华为数字技术有限公司 | Abnormal voice classifying method and device |
| CN107393555A (en) * | 2017-07-14 | 2017-11-24 | 西安交通大学 | A kind of detecting system and detection method of low signal-to-noise ratio abnormal sound signal |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040083094A1 (en) * | 2002-10-29 | 2004-04-29 | Texas Instruments Incorporated | Wavelet-based compression and decompression of audio sample sets |
| JP6344383B2 (en) * | 2013-04-26 | 2018-06-20 | 日本電気株式会社 | Behavior analysis device, behavior analysis method, and behavior analysis program |
-
2017
- 2017-12-28 CN CN201711462639.1A patent/CN108182950B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1207170A (en) * | 1996-09-13 | 1999-02-03 | 株式会社东芝 | Unsteady signal analyzer and medium for recording unsteady signal analyzer program |
| CN103366734A (en) * | 2012-03-31 | 2013-10-23 | 佳能株式会社 | Method and device for checking voice recognition results, voice recognition system and audio monitoring system |
| CN106228979A (en) * | 2016-08-16 | 2016-12-14 | 重庆大学 | A kind of abnormal sound in public places feature extraction and recognition methods |
| CN106683687A (en) * | 2016-12-30 | 2017-05-17 | 杭州华为数字技术有限公司 | Abnormal voice classifying method and device |
| CN107393555A (en) * | 2017-07-14 | 2017-11-24 | 西安交通大学 | A kind of detecting system and detection method of low signal-to-noise ratio abnormal sound signal |
Non-Patent Citations (5)
| Title |
|---|
| Fault diagnosis of motorized spindle via modified empirical wavelet transform-kernel PCA and optimized support vector machine;Fei Chen et al;《JVE journals》;20170630;全文 * |
| FFT and Wavelet-Based Feature Extraction for Acoustic Audio Classification;A.K.M Fazlul Haque;《International Journal of Advance Innovations, Thoughts & Ideas》;20121231;全文 * |
| 公共场所典型异常声音的特征提取;栾少文等;《计算机工程》;20100430;第36卷(第7期);全文 * |
| 改进的ESMD用于公共场所异常声音特征提取;李伟红等;《仪器仪表学报》;20161130;第37卷(第11期);全文 * |
| 智能监控系统中公共场所异常声音检测的研究;栾少文;《中国优秀硕士学位论文全文数据库 信息科技辑》;20091215(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108182950A (en) | 2018-06-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | A deep neural network for time-domain signal reconstruction | |
| CN109616140B (en) | Abnormal sound analysis system | |
| CN106846803B (en) | Traffic event detection device and method based on audio frequency | |
| Valero et al. | Gammatone cepstral coefficients: Biologically inspired features for non-speech audio classification | |
| CN102664006B (en) | Abnormal voice detecting method based on time-domain and frequency-domain analysis | |
| CN106683687B (en) | Abnormal sound classification method and device | |
| Sharan et al. | Cochleagram image feature for improved robustness in sound recognition | |
| CN111785285A (en) | Voiceprint recognition method for home multi-feature parameter fusion | |
| CN112017682B (en) | Single-channel voice simultaneous noise reduction and reverberation removal system | |
| CN111933186B (en) | Method, device and system for fault identification of on-load tap-changer | |
| CN111489763B (en) | GMM model-based speaker recognition self-adaption method in complex environment | |
| Nossier et al. | Mapping and masking targets comparison using different deep learning based speech enhancement architectures | |
| CN110909827A (en) | Noise reduction method suitable for fan blade sound signals | |
| Lim et al. | Harmonic and percussive source separation using a convolutional auto encoder | |
| CN108182950B (en) | Improved method for decomposing and extracting abnormal sound characteristics of public places through empirical wavelet transform | |
| CN116778956A (en) | Transformer acoustic feature extraction and fault identification method | |
| CN108922514A (en) | A kind of robust features extracting method based on low frequency logarithmic spectrum | |
| CN109920447B (en) | Recording fraud detection method based on adaptive filter amplitude phase characteristic extraction | |
| Huang et al. | Dccrgan: Deep complex convolution recurrent generator adversarial network for speech enhancement | |
| CN111785262B (en) | Speaker age and gender classification method based on residual error network and fusion characteristics | |
| Fraile et al. | Mfcc-based remote pathology detection on speech transmitted through the telephone channel-impact of linear distortions: Band limitation, frequency response and noise | |
| Zhang et al. | A pairwise algorithm for pitch estimation and speech separation using deep stacking network | |
| CN115293205A (en) | Anomaly detection method, self-encoder model training method and electronic equipment | |
| CN115171021A (en) | GPU-based passenger real-time interactive processing method and system | |
| CN114664311A (en) | Memory network enhanced variational inference silent attack detection method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |