CN109408660A - A method of the music based on audio frequency characteristics is classified automatically - Google Patents
A method of the music based on audio frequency characteristics is classified automatically Download PDFInfo
- Publication number
- CN109408660A CN109408660A CN201811012539.3A CN201811012539A CN109408660A CN 109408660 A CN109408660 A CN 109408660A CN 201811012539 A CN201811012539 A CN 201811012539A CN 109408660 A CN109408660 A CN 109408660A
- Authority
- CN
- China
- Prior art keywords
- music
- frequency characteristics
- audio frequency
- training
- test set
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Auxiliary Devices For Music (AREA)
Abstract
The invention discloses a kind of methods that the music based on audio frequency characteristics is classified automatically, comprising: obtains the music building music training set and music test set of different classifications;Audio feature extraction is carried out to the music in the music and the music test set in the music training set respectively, and obtains the musical features collection of music training set and the musical features collection of music test set;The musical features collection of the music training set as training sample and is subjected to neural metwork training, neural network after being trained, and using the musical features collection of the music test set as test sample and the neural network after training is tested, and adjust neural network parameter.The present invention is from the audio frequency characteristics of music, the audio frequency characteristics of musical specific property are selected preferably to express, solve the problems, such as that audio frequency characteristics are single, effective data basis is provided to classify automatically, the accuracy that music is classified automatically is improved, and neural metwork training is carried out to audio frequency characteristics on this basis.
Description
Technical field
The present invention relates to field of computer technology, method that especially a kind of music based on audio frequency characteristics is classified automatically.
Background technique
With the fast development of multimedia technology and network technology, a large amount of music data is caused to circulate on the net, and more
Music data in media database is in explosive growth.But can the value of large-scale music data effective with user
The content that ground browses music data is closely bound up.Therefore, people are badly in need of the effective automatic management of research and retrieval music data
Method cope with such development trend, and in recent years, the automatic classification technology and retrieval technique of music has become research heat
Topic.
The classification automatically of music is the process of a pattern-recognition, including audio feature extraction and automatic classification two are substantially
Process.The audio frequency characteristics extracted in the prior art are relatively simple, and some lays particular emphasis on tone color, and some lays particular emphasis on rhythm, are not enough to complete
The expression musical specific property of face entirety faces large-scale music data, cannot be finely accurately right in subsequent automatic classification
Music is classified.
Summary of the invention
In order to overcome above-mentioned defect in the prior art, the present invention provides a kind of music based on audio frequency characteristics and classifies automatically
Method selected the audio frequency characteristics that can preferably express musical specific property from the audio frequency characteristics of music, solve audio spy
Single problem is levied, effective data basis is provided for automatic classification, improves the accuracy that music is classified automatically, and herein
On the basis of to audio frequency characteristics carry out neural metwork training.
To achieve the above object, the present invention uses following technical scheme, comprising:
A method of the music based on audio frequency characteristics is classified automatically, comprising the following steps:
S1, obtains the music building music collections of different classifications, the music collections include for trained music training set and
For the music test set of test, and music training set and music test set include the music of different classifications;
S2 carries out audio frequency characteristics to the music in the music and the music test set in the music training set respectively and mentions
It takes, obtains the audio frequency characteristics of the music in the music training set, and constitute the music of music training set according to this audio frequency characteristics
Feature set;And the audio frequency characteristics of the music in the music test set are obtained, and constitute music test according to this audio frequency characteristics
The musical features collection of collection;
The audio frequency characteristics include: Mel Cepstral Frequency Coefficients MFCC, music energy, music rhythm;
The musical features collection of the music training set as training sample and is carried out neural metwork training, is instructed by S3
Neural network after white silk;And using the musical features collection of the music test set as test sample and to the neural network after training
It is tested.
In step S1, the mode classification of the music of the different classifications includes: prevalence, R&B, jazz, electronics, folk rhyme, shakes
Rolling, a Chinese musical telling, rural area.
In step S2, using MATLB to the music in the music and the music test set in the music training set into
Row audio feature extraction.
In step S2, the extracting mode of the Mel Cepstral Frequency Coefficients MFCC are as follows: pre-add is carried out to the music signal of music
Weight, framing, adding window, Fast Fourier Transform (FFT), Mel filtering, logarithm operation, discrete cosine transform, obtain the Meier scramble of 12 dimensions
Spectral coefficient, and second differnce is carried out to the Mel Cepstral Frequency Coefficients of 12 dimension, the Mel Cepstral Frequency Coefficients of 24 dimensions are obtained, by plum
That cepstral coefficients are used as a characteristic parameter per the numerical value on one-dimensional, obtain 24 according to the Mel Cepstral Frequency Coefficients of 24 dimensions
A characteristic parameter.
In step S2, the extracting mode of the music energy are as follows: preemphasis is carried out to the music signal of music, framing, is added
Window, Fast Fourier Transform (FFT), and music energy is obtained according to the Fourier spectrum figure after Fast Fourier Transform (FFT);The music energy
Amount includes 2 characteristic parameters, the respectively mean value of the maximum amplitude in Fourier spectrum figure and amplitude.
In step S2, the extracting mode of the music rhythm are as follows: extract the bat number of the every five seconds of music, and according to every five seconds
Bat number obtains music rhythm;The music rhythm includes 2 characteristic parameters, the respectively maximum bat in every five seconds bat number
Several and every five seconds bat number mean value.
In step S3, the neural metwork training is to be trained using BP neural network algorithm;The BP network neural
Algorithm: input training sample carries out adjusting training repeatedly using weight and threshold value of the back-propagation algorithm to BP neural network,
The difference between output vector and Mean Vector is set to be less than the 10% of Mean Vector;The output vector is that the prediction of music is classified
Value;The Mean Vector is the actual classification value of music.
In step S3, the BP neural network after training is tested to obtain the classification accuracy of the BP neural network,
Number of levels, circulation if the classification accuracy less than 85%, needs to readjust neural network parameter, to BP neural network
Number, learning rate, target error are modified;According to the quantity and music test set of accurate music of classifying in music test set
In the quantity of all music obtain the classification accuracy, the number for accurate music of classifying in classification accuracy=music test set
The quantity of all music in amount/music test set.
The present invention has the advantages that
(1) audio frequency characteristics of the extracted music of the present invention include 28 characteristic parameters, more comprehensively whole to music spy
Property is expressed, to accurately classify to music.
(2) when the present invention extracts Mel Cepstral Frequency Coefficients MFCC, second differnce has been carried out, it is clearer to represent
Music variation characteristic keeps the extraction of the tamber characteristic of music more accurate.
(3) present invention is trained the automatic classification of music using BP neural network algorithm, and leads to during the test
Cross constantly adjustment neural network parameter value mode, to the number of levels of neural network, cycle-index, learning rate, target error into
Row modification, makes classification accuracy be higher than 85%, to guarantee the classifying quality that music is classified automatically.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
As shown in Figure 1, a method of the music based on audio frequency characteristics is classified automatically, comprising the following steps:
S1 obtains the music of different classifications according to the mainstream mode classification of current music, constitutes music collections;It will be in music collections
80% as the music training set for training, by 20% in music collections as the music test set for being used for test, and institute
State include in music training set and the music test set different classifications music.
The mainstream mode classification includes: prevalence, R&B, jazz, electronics, folk rhyme, rock and roll, a Chinese musical telling, rural area.
S2 respectively carries out the music in the music and the music test set in the music training set using MATLAB
Audio feature extraction obtains the audio frequency characteristics of the music in the music training set, and constitutes music instruction according to this audio frequency characteristics
Practice the musical features collection of collection;And the audio frequency characteristics of the music in the music test set are obtained, and according to this audio frequency characteristics structure
At the musical features collection of music test set;
The audio frequency characteristics include: Mel Cepstral Frequency Coefficients MFCC, music energy, music rhythm.
Wherein, the extracting mode of the Mel Cepstral Frequency Coefficients MFCC are as follows: preemphasis is carried out to the music signal of music, is divided
Frame, adding window, Fast Fourier Transform (FFT), Mel filtering, logarithm operation, discrete cosine transform, obtain the Mel-cepstral system of 12 dimensions
Number carries out second differnce to the Mel Cepstral Frequency Coefficients of 12 dimension, obtains to more clearly from represent music variation characteristic
To the Mel Cepstral Frequency Coefficients of 24 dimensions, Mel Cepstral Frequency Coefficients are regard as a characteristic parameter, root per the numerical value on one-dimensional
24 characteristic parameters are obtained according to the Mel Cepstral Frequency Coefficients of 24 dimensions.
The extracting mode of the music energy are as follows: preemphasis, framing, adding window, quick Fu are carried out to the music signal of music
In leaf transformation, and music energy is obtained according to the Fourier spectrum figure after Fast Fourier Transform (FFT);The music energy includes 2
Characteristic parameter, the respectively mean value of the maximum amplitude in Fourier spectrum figure and amplitude.
The extracting mode of the music rhythm are as follows: extract the bat number of the every five seconds of music, and obtained according to every five seconds bat number
To music rhythm;The music rhythm includes 2 characteristic parameters, maximum bat number and every five seconds respectively in every five seconds bat number
The mean value of bat number.
The audio frequency characteristics of a piece of music include 28 characteristic parameters, are respectively obtained by the Mel Cepstral Frequency Coefficients of 24 dimensions
24 characteristic parameters, 2 characteristic parameters for indicating music energy, 2 characteristic parameters for indicating music rhythm.
The musical features collection of the music training set as training sample and is carried out neural metwork training, is instructed by S3
BP neural network after white silk;And the BP mind using the musical features collection of the music test set as test sample, after test training
Classification accuracy through network, it is right if the classification accuracy of BP neural network less than 85%, readjusts neural network parameter
The number of levels of BP neural network, cycle-index, learning rate, target error are modified;If the classification accuracy of BP neural network
More than or equal to 92.5%, then it represents that test effect is good, i.e., the BP neural network after the training has good classification accurate
Property.
The neural metwork training is to be trained using BP neural network algorithm;The BP network neural algorithm: input
Training sample carries out adjusting training repeatedly to the weight and threshold value of BP neural network using back-propagation algorithm, make output to
Difference between amount and Mean Vector is less than the 10% of Mean Vector;The output vector is the prediction classification value of music, is BP
The classification value of neural network prediction;The Mean Vector is the actual classification value of music, is the Accurate classification value of music training set.
When carrying out the training and test of BP neural network, one is added before 28 characteristic parameters of the audio frequency characteristics
For the characteristic parameter of mark, that is, the audio frequency characteristics after increasing the characteristic parameter for mark include 29 characteristic parameters, respectively
2 characteristic parameters, expression music for 24 characteristic parameters, expression music energy that are obtained by the Mel Cepstral Frequency Coefficients of 24 dimensions
2 characteristic parameters of rhythm, 1 characteristic parameter for mark;The characteristic parameter for mark is to identify music not
With classification, prevalence, R&B, jazz, electronics, folk rhyme, rock and roll, a Chinese musical telling, rural area are respectively indicated with 1,2,3,4,5,6,7,8.
When carrying out the test of BP neural network, input parameter is the musical features collection of music test set, and output parameter is
Music assorting corresponding to music in music test set, and the output parameter of test is also used into 1,2,3,4,5,6,7,8 difference
Indicate prevalence, R&B, jazz, electronics, folk rhyme, rock and roll, a Chinese musical telling, rural area.If in the audio frequency characteristics of certain song for mark
Characteristic parameter is 1, then output parameter of the audio frequency characteristics of the song after BP neural network is tested also is to answer 1, i.e., according to defeated
Out parameter whether unanimously come with the characteristic parameter for mark the BP neural network after training of judgement classification it is whether accurate, if one
It causes, then classification is accurate;It otherwise is classification inaccuracy.
It is obtained according to the quantity of all music in the quantity for accurate music of classifying in music test set and music test set
The classification accuracy owns in the quantity/music test set for accurate music of classifying in classification accuracy=music test set
The quantity of music.
The present invention has selected the audio frequency characteristics that can preferably express musical specific property from the audio frequency characteristics of music, solves
Audio frequency characteristics single problems provide effective data basis for automatic classification, improve music classify automatically it is accurate
Property, and neural metwork training is carried out to audio frequency characteristics on the basis of this audio feature extraction, the automatic classification of music is made
It is further to improve, to construct the higher music automatic categorizer of classification accuracy.The present invention is according to 24 Jan Vermeer scrambles
Spectral coefficient obtains 24 characteristic parameters of audio frequency characteristics, equally, according to Mel Cepstral Frequency Coefficients extracting mode and carry out difference
More high-dimensional Mel Cepstral Frequency Coefficients can be obtained, to obtain the more features parameter of audio frequency characteristics in processing.
The above is only the preferred embodiments of the invention, are not intended to limit the invention creation, all in the present invention
Made any modifications, equivalent replacements, and improvements etc., should be included in the guarantor of the invention within the spirit and principle of creation
Within the scope of shield.
Claims (8)
1. a kind of method that the music based on audio frequency characteristics is classified automatically, which comprises the following steps:
S1 obtains the music building music collections of different classifications, and the music collections include for trained music training set and being used for
The music test set of test, and music training set and music test set include the music of different classifications;
S2 carries out audio feature extraction to the music in the music and the music test set in the music training set respectively,
The audio frequency characteristics of the music in the music training set are obtained, and constitute the musical features of music training set according to this audio frequency characteristics
Collection;And the audio frequency characteristics of the music in the music test set are obtained, and constitute music test set according to this audio frequency characteristics
Musical features collection;
The audio frequency characteristics include: Mel Cepstral Frequency Coefficients MFCC, music energy, music rhythm;
The musical features collection of the music training set as training sample and is carried out neural metwork training, after being trained by S3
Neural network;And using the musical features collection of the music test set as test sample and the neural network after training is carried out
Test.
2. the method that a kind of music based on audio frequency characteristics according to claim 1 is classified automatically, which is characterized in that step
In S1, the mode classification of the music of the different classifications includes: prevalence, R&B, jazz, electronics, folk rhyme, rock and roll, a Chinese musical telling, rural area.
3. the method that a kind of music based on audio frequency characteristics according to claim 1 is classified automatically, which is characterized in that step
In S2, audio frequency characteristics are carried out to the music in the music and the music test set in the music training set using MATLB and are mentioned
It takes.
4. the method that a kind of music based on audio frequency characteristics according to claim 3 is classified automatically, which is characterized in that step
In S2, the extracting mode of the Mel Cepstral Frequency Coefficients MFCC are as follows: to the music signal of music carry out preemphasis, framing, adding window,
Fast Fourier Transform (FFT), Mel filtering, logarithm operation, discrete cosine transform, obtain the Mel Cepstral Frequency Coefficients of 12 dimensions, and to institute
The Mel Cepstral Frequency Coefficients for stating 12 dimensions carry out second differnce, the Mel Cepstral Frequency Coefficients of 24 dimensions are obtained, by Mel Cepstral Frequency Coefficients
Be used as a characteristic parameter per the numerical value on one-dimensional, obtain 24 characteristic parameters according to the Mel Cepstral Frequency Coefficients of 24 dimensions.
5. the method that a kind of music based on audio frequency characteristics according to claim 3 is classified automatically, which is characterized in that step
In S2, the extracting mode of the music energy are as follows: preemphasis, framing, adding window, fast Fourier are carried out to the music signal of music
Transformation, and music energy is obtained according to the Fourier spectrum figure after Fast Fourier Transform (FFT);The music energy includes 2 features
Parameter, the respectively mean value of the maximum amplitude in Fourier spectrum figure and amplitude.
6. the method that a kind of music based on audio frequency characteristics according to claim 3 is classified automatically, which is characterized in that step
In S2, the extracting mode of the music rhythm are as follows: extract the bat number of the every five seconds of music, and sound is obtained according to every five seconds bat number
Happy rhythm;The music rhythm includes 2 characteristic parameters, maximum bat number and every five seconds bat respectively in every five seconds bat number
Several mean values.
7. the method that a kind of music based on audio frequency characteristics according to claim 1 is classified automatically, which is characterized in that step
In S3, the neural metwork training is to be trained using BP neural network algorithm;The BP network neural algorithm: input training
Sample carries out adjusting training repeatedly using weight and threshold value of the back-propagation algorithm to neural network, makes output vector and phase
The difference hoped between vector is less than the 10% of Mean Vector;The output vector is the prediction classification value of music;It is described it is expected to
Amount is the actual classification value of music.
8. the method that a kind of music based on audio frequency characteristics according to claim 1 is classified automatically, which is characterized in that step
In S3, the BP neural network after training is tested to obtain the classification accuracy of the BP neural network, if classification accuracy
It less than 85%, then needs to readjust neural network parameter, to the number of levels of BP neural network, cycle-index, learning rate, target
Error is modified;According to the quantity of all music in the quantity for accurate music of classifying in music test set and music test set
The classification accuracy is obtained, in the quantity/music test set for accurate music of classifying in classification accuracy=music test set
The quantity of all music.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811012539.3A CN109408660B (en) | 2018-08-31 | 2018-08-31 | Music automatic classification method based on audio features |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811012539.3A CN109408660B (en) | 2018-08-31 | 2018-08-31 | Music automatic classification method based on audio features |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109408660A true CN109408660A (en) | 2019-03-01 |
CN109408660B CN109408660B (en) | 2021-08-10 |
Family
ID=65463706
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811012539.3A Active CN109408660B (en) | 2018-08-31 | 2018-08-31 | Music automatic classification method based on audio features |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109408660B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110427518A (en) * | 2019-08-12 | 2019-11-08 | 央视国际网络无锡有限公司 | A kind of short Video Music recommended method |
CN110555457A (en) * | 2019-07-24 | 2019-12-10 | 天津大学 | Engine lubricating oil wear signal characteristic processing system and method |
CN113012713A (en) * | 2021-03-02 | 2021-06-22 | 哈尔滨理工大学 | Music genre classification method based on logistic regression algorithm in machine learning |
CN113658611A (en) * | 2021-08-11 | 2021-11-16 | 华南农业大学 | CNN-based Guangdong drama genre classification and identification method |
CN113793580A (en) * | 2021-08-31 | 2021-12-14 | 云境商务智能研究院南京有限公司 | Music genre classification method based on deep learning |
CN114036341A (en) * | 2022-01-10 | 2022-02-11 | 腾讯科技(深圳)有限公司 | Music tag prediction method and related equipment |
CN114352486A (en) * | 2021-12-31 | 2022-04-15 | 西安翔迅科技有限责任公司 | Wind turbine generator blade audio fault detection method based on classification |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040231498A1 (en) * | 2003-02-14 | 2004-11-25 | Tao Li | Music feature extraction using wavelet coefficient histograms |
CN103186527A (en) * | 2011-12-27 | 2013-07-03 | 北京百度网讯科技有限公司 | System for building music classification model, system for recommending music and corresponding method |
CN108257614A (en) * | 2016-12-29 | 2018-07-06 | 北京酷我科技有限公司 | The method and its system of audio data mark |
CN108305619A (en) * | 2017-03-10 | 2018-07-20 | 腾讯科技(深圳)有限公司 | Voice data collection training method and apparatus |
-
2018
- 2018-08-31 CN CN201811012539.3A patent/CN109408660B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040231498A1 (en) * | 2003-02-14 | 2004-11-25 | Tao Li | Music feature extraction using wavelet coefficient histograms |
CN103186527A (en) * | 2011-12-27 | 2013-07-03 | 北京百度网讯科技有限公司 | System for building music classification model, system for recommending music and corresponding method |
CN108257614A (en) * | 2016-12-29 | 2018-07-06 | 北京酷我科技有限公司 | The method and its system of audio data mark |
CN108305619A (en) * | 2017-03-10 | 2018-07-20 | 腾讯科技(深圳)有限公司 | Voice data collection training method and apparatus |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110555457A (en) * | 2019-07-24 | 2019-12-10 | 天津大学 | Engine lubricating oil wear signal characteristic processing system and method |
CN110427518A (en) * | 2019-08-12 | 2019-11-08 | 央视国际网络无锡有限公司 | A kind of short Video Music recommended method |
CN113012713A (en) * | 2021-03-02 | 2021-06-22 | 哈尔滨理工大学 | Music genre classification method based on logistic regression algorithm in machine learning |
CN113658611A (en) * | 2021-08-11 | 2021-11-16 | 华南农业大学 | CNN-based Guangdong drama genre classification and identification method |
CN113793580A (en) * | 2021-08-31 | 2021-12-14 | 云境商务智能研究院南京有限公司 | Music genre classification method based on deep learning |
CN113793580B (en) * | 2021-08-31 | 2024-05-24 | 云境商务智能研究院南京有限公司 | Music genre classification method based on deep learning |
CN114352486A (en) * | 2021-12-31 | 2022-04-15 | 西安翔迅科技有限责任公司 | Wind turbine generator blade audio fault detection method based on classification |
CN114036341A (en) * | 2022-01-10 | 2022-02-11 | 腾讯科技(深圳)有限公司 | Music tag prediction method and related equipment |
CN114036341B (en) * | 2022-01-10 | 2022-03-29 | 腾讯科技(深圳)有限公司 | Music tag prediction method and related equipment |
Also Published As
Publication number | Publication date |
---|---|
CN109408660B (en) | 2021-08-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109408660A (en) | A method of the music based on audio frequency characteristics is classified automatically | |
KR101269296B1 (en) | Neural network classifier for separating audio sources from a monophonic audio signal | |
CN103218608B (en) | Network violent video identification method | |
CN108922560A (en) | A kind of city noise recognition methods based on interacting depth neural network model | |
Huang et al. | Intelligent feature extraction and classification of anuran vocalizations | |
CN109360554A (en) | A kind of language identification method based on language deep neural network | |
Jiang et al. | An Improved Speech Segmentation and Clustering Algorithm Based on SOM and K‐Means | |
CN111653289A (en) | Playback voice detection method | |
CN108831506A (en) | Digital audio based on GMM-BIC distorts point detecting method and system | |
CN102201237A (en) | Emotional speaker identification method based on reliability detection of fuzzy support vector machine | |
CN106649513B (en) | Audio data clustering method based on spectral clustering | |
CN108538312A (en) | Digital audio based on bayesian information criterion distorts a method for automatic positioning | |
CN108766464A (en) | Digital audio based on mains frequency fluctuation super vector distorts automatic testing method | |
CN104992708B (en) | Specific audio detection model generation in short-term and detection method | |
Wu et al. | A feature selection algorithm of music genre classification based on ReliefF and SFS | |
Xiao et al. | AMResNet: An automatic recognition model of bird sounds in real environment | |
CN111508528B (en) | No-reference audio quality evaluation method and device based on natural audio statistical characteristics | |
CN110739006A (en) | Audio processing method and device, storage medium and electronic equipment | |
CN110288976A (en) | Data screening method, apparatus and intelligent sound box | |
CN111554273B (en) | Method for selecting amplified corpora in voice keyword recognition | |
CN113936667A (en) | Bird song recognition model training method, recognition method and storage medium | |
CN114067788A (en) | Guangdong opera vocal cavity classification method based on combination of CNN and LSTM | |
CN110689875A (en) | Language identification method and device and readable storage medium | |
CN111259861A (en) | Underwater acoustic communication signal modulation mode identification method under data set unbalanced condition | |
Liu et al. | Birdcall identification using mel-spectrum based on ResNeSt50 model |
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 |