Home » Source Code » Speaker Recognition Matlab Code

Speaker Recognition Matlab Code

casiopia
2014-10-13 20:41:00
The author
View(s):
Download(s): 35
Point (s): 1 
Category Category:
MatlabMatlab MatlabMatlab

Description

Speaker Recognition


Documentation contains:


1. Introduction
2. Previous works
3. Theoretical concepts
4. Algorithm
5. Experimentation
6. Results
7. Plots and Graphs
8. Accuracy values
9. Future Scope
10. References


1.Introduction: 
In the domain of Speaker recognition the selected research paper to be implemented is “Voice
Recognition using HMM with MFCC for Secure ATM” by “Shumaila Iqbal, Tahira Mahboob and
Malik Sikandar Hayat Khiyal”.

In this paper, voice sample is observed with MFCC for extracting acoustic features and then used to trained HMM parameters through forward backward algorithm which lies under HMM and finally the computed log likelihood from training is stored to database. It will recognize the speaker by comparing the log value from the database against the PIN code. It is implemented in Matlab 7.0 environment and showing 86.67% results as correct acceptance and correct rejections with the error rate of 13.33%.


2.Previous works: 
In Early research Shi-Huang Chen and Yu-Ren Luo [1] used MFCC to extract features. Trained and
recognized using SVM (support Vector Machine). SVM technique based on two class classifiers.
Decision in binary form was introduced. Discrimination of claimed speaker and imposer by +1 and -1.
Results averaged accuracy rate to 95.1% with EER of 0.0%.

In another research, Lindasalwa Muda, Mumtaj Begam and I. Elamvazuthi [2] also used MFCC for voice
features extraction. Trained and recognized using DTW. DTW (dynamic time Warping) a technique
based on dynamic programing, used for recognition process. Measures similarity between two time
series which may vary in time or speed.

One other research by Ibrahim Patel and Dr. Y. Srinivas Rao [3] introduced sub band coding. The
integration of MFCC with sub band coding increases its efficiency and accurate classification as
compared to MFCC separately. The two features of MFCC and integrated sub band decomposition
with MFCC are used in HMM to train and recognize the speaker.




3.Theoretical concepts:
Feature Extraction:
Feature vector calculation using the feature ‘Mel frequency cepstral coefficients’. Algorithm to calculate
mfcc is as below:

 

mfcc


Classification:
Euclidian distance, dE = dist(x,y) = sqrt [(x1-y1)2 + (x2-y2)2 +……+(xN-yN)2]
Where feature vector x = (x1,x2,x3,…..,xN ) and feature vector y = (y1,y2,y3,…..,yN ) 
FV_mean is the mean of training vectors. For each class testing sample is compared to FV_mean to
get the distance.

dE = dist(FVectorTest, FV_mean_class1)
Min(dE) decides the match for proper class.
 
4.Algorithm:
1. DataSet of voice samples is collected.
2. Training Dataset is read through ‘wavread’ function.
3. Mfcc feature vector is calculated on Training Dataset.[5]
4. Testing Dataset is read through ‘wavread’ function.
5. Mfcc feature vector is calculated on Testing Dataset.[5] 
6. Mean feature vector of each class is calculated.
7. Euclidean distance is calculated to classify the testing feature vectors.
8. Minimum distance is taken as the deciding factor of classification.


5.Experimentation:
Five classes of training and testing data are used in this experiment within which 3 are for training and
2 for testing for each class. Sampling frequency used for recording is 44 KHz and each sound clip is of
4 sec length. Noisy signals are taken and without applying any filter, MFCC has been calculated. While
calculating MFCC first 15 coefficients are taken into consideration. Euclidean distance is used for
classification.



6.Result: 
Class1

Euclidean distance from

TestSample1

TestSample2

Class1 mean

243.7097

345.2720

Class2 mean

319.4568

531.7993

Class3 mean

248.0988

462.0911

Class4 mean

784.1279

797.2653

Class5 mean

678.4059

568.0027



Class2

Euclidean distance from

TestSample3

TestSample4

Class1 mean

386.4527 

363.1666

Class2 mean

251.0477 

224.8126

Class3 mean

349.3757 

324.8985

Class4 mean

741.0495 

733.5317

Class5 mean

713.3305 

703.1649



Class3

Euclidean distance from

TestSample5

TestSample6

Class1 mean

327.1036 

307.1795

Class2 mean

336.2672 

325.9956

Class3 mean

161.5853 

169.4514

Class4 mean

 822.4920 

791.3391

Class5 mean

776.9189 

726.4291




Class4

Euclidean distance from

TestSample7

TestSample8

Class1 mean

686.6294 

624.4262

Class2 mean

640.9071 

631.8815

Class3 mean

689.3472 

722.6730

Class4 mean

523.9794 

666.5829

Class5 mean

718.9898 

633.1017




Class5

Euclidean distance from

TestSample9

TestSample10

Class1 mean

570.9322 

679.5080

Class2 mean

698.2867 

784.6366

Class3 mean

712.7408 

823.7812

Class4 mean

792.5957 

814.5246

Class5 mean

437.1480 

428.5042







7.Plots and graphs:


  
plot

 
8.Accuracy values:
Accuracy obtained: 90%


9.Future Scope:
The application software Matlab7.0 is taking a couple of minutes to process the dataset. This could
cause serious problem while working with a really large dataset. There is scope for further
improvements.


10.References:
[1] Shi-Huang Chen and Yu-Ren Luo,” Speaker Verification Using MFCC and Support Vector Machine”,
Proceedings of the International Multi Conference of Engineers and Computer Scientists, Vol I, IMECS
2009, March 2009 

[2] Lindasalwa Muda, Mumtaj Begam and I. Elamvazuthi, “Voice Recognition Algorithms using Mel
Frequency Cepstral Coefficient (MFCC) and Dynamic Time Warping (DTW) Techniques” Journal Of
Computing, Volume 2, Issue 3, March 2010

[3] Ibrahim Patel and Dr. Y. Srinivas Rao, “Speech Recognition Using Hmm With Mfcc- An Analysis
Using Frequency Spectral Decomposing Technique”, an International Journal (SIPIJ) Vol.1, No.2,
December 2010 

[4] Shumaila Iqbal, Tahira Mahboob and Malik Sikandar Hayat Khiyal ,“Voice Recognition System
using HMM with MFCC for secure ATM”, IJCSI International Journal of Computer Science Issues,
Vol. 8, Issue 6, No 3, November 2011,Page 297-302

[5] http://www.mathworks.in/matlabcentral/fileexchange/23119-mfcc/content/kannumfcc.m



Sponsored links

File list

Tips: You can preview the content of files by clicking file names^_^
Name Size Date
CalcMfcc.m2.17 kB03-02-14 16:13
classify.m4.87 kB06-07-14 15:56
<1>0.00 B53% 04-07-14
<1>0.00 B56% 04-07-14
<1>0.00 B54% 04-07-14
<1>0.00 B54% 04-07-14
<1>0.00 B52% 04-07-14
<2>0.00 B54% 06-07-14
<2>0.00 B55% 06-07-14
<2>0.00 B56% 06-07-14
<2>0.00 B55% 06-07-14
<2>0.00 B55% 06-07-14
<3>0.00 B53% 04-07-14
<3>0.00 B56% 04-07-14
<3>0.00 B54% 04-07-14
<3>0.00 B54% 04-07-14
<3>0.00 B52% 04-07-14
<4>0.00 B54% 06-07-14
<4>0.00 B55% 06-07-14
<4>0.00 B56% 06-07-14
<4>0.00 B55% 06-07-14
<4>0.00 B55% 06-07-14
<5>0.00 B53% 04-07-14
<5>0.00 B56% 04-07-14
<5>0.00 B54% 04-07-14
<5>0.00 B54% 04-07-14
<5>0.00 B52% 04-07-14
dist.m200.00 B27-04-14 17:09
distance.fig74.50 kB27-04-14 19:55
frq2mel.m2.73 kB23-12-13 14:43
mel2frq.m2.72 kB23-12-13 14:44
melbankm.m11.11 kB23-12-13 14:41
plots.m1.11 kB27-04-14 19:44
process_audio.m53.00 B27-04-14 23:44
Project_Documentation.pdf295.41 kB27-04-14 22:38
start.m219.00 B06-07-14 15:28
test.m232.00 B27-04-14 19:11
train.m217.00 B27-04-14 19:11
<data>0.00 B10-10-14 18:57
<SpeakerRecognitionMatlabCode>0.00 B10-10-14 18:59
...
Sponsored links

Comments

(Add your comment, get 0.1 Point)
Minimum:15 words, Maximum:160 words
st2013
2014-12-24

很好的代码,讲的很详细,很有参考价值!

为你停留
2015-01-23

对matlab比较感兴趣,有空的话会关注的

nikolas_xie
2015-03-27

我自己之前做了一个基于DTW算法的最简单的说话人识别,没有加滤波器,只是用最简单算法,识别率不高。学习一下这份作品。

evansis
2015-04-11

this really helped me. Your code really helped a lot.

刺客123
2015-04-18

挺好的,正在学习模式识别,对于一些知识还是不怎么懂

luninglang
2015-04-24

学习当中,学习中,似乎很多人都在学这个代码编程

mingming713
2015-05-03

是个不错的项目,代码也焊好。非常感谢楼主

惢瑞
2015-05-04

讲解的是挺详细的,准备下载看看 但是cf币不够

lanzhusmile
2015-05-10

很不错,学习学习的好资料,可以增加很多的知识

liuzzliu
2015-05-11

正在学习中,很有帮助和价值,谢谢

dongnob
2015-05-18

感谢lz的分享!应该非常有用!谢谢!最近正在做一个交互方面的软件,希望能得到帮助!

shuimove
2015-05-24

里面包含有测试的语音文件,代码详细,还有说明的pdf,推荐下载

一撧版小女子
2016-03-17

最近正在研究这个课题,希望能得到帮助,对matlab也不熟悉。不过还是很感谢分享,大家共同进步!

ckldan520
2016-03-21

好像很好的样子,希望能够帮助自己得到锻炼

静马酱
2016-04-14

我也正在学习使用MATLAB声纹识别方面内容,前来借鉴学习

康欣红
2016-06-07

想下载一些代码,但是积分有限,如何赚积分呢???

rojaparveen
2017-02-05

thank u

拾四喵777
2017-02-12

最近在学习说话人识别的内容,想看一下代码,下载看看

dan66
2017-02-28

cool

cnbluehwa
2017-03-04

本科毕设课题,正在艰难的学习中,看到这真是太给力了。

Speaker Recognition Matlab Code (4.95 MB)

Need 1 Point(s)
Your Point (s)

Your Point isn't enough.

Get 22 Point immediately by PayPal

Point will be added to your account automatically after the transaction.

More(Debit card / Credit card / PayPal Credit / Online Banking)

Submit your source codes. Get more Points

LOGIN

Don't have an account? Register now
Need any help?
Mail to: support@codeforge.com

切换到中文版?

CodeForge Chinese Version
CodeForge English Version

Where are you going?

^_^"Oops ...

Sorry!This guy is mysterious, its blog hasn't been opened, try another, please!
OK

Warm tip!

CodeForge to FavoriteFavorite by Ctrl+D