TWO-STAGE DECISION MAKING ALGORITHM FOR SPEAKER VERIFICATION WITH TRAINING SET OPTIMIZATION

Abstract

   In this paper, a two-stage decision making algorithm is proposed for the task of speaker verification. This two-stage algorithm aims to eliminate the first-stage qualifying impostors by the help of impostor-resistant structure in the second stage. First, a baseline system is formed using mel-frequency cepstral coefficients (MFCC) as features and, a radial basis function (RBF) neural network for speaker modelling. Then, the investigations have been realized for optimizing the training set by means of two issues: (1) the ratio of impostor features to genuine speaker features, (2) the ratio of same gender features to opposite gender features (in respect of the genuine speaker) within the impostor speakers’ set. Last, the two-stage decision making algorithm is presented, and the performance enhancement provided by the two-stage system is given with the test results.

Keywords

• Speaker verification,Training set optimization,RBF neural network,MFCC,Cohort

KONUŞMACI DOĞRULAMA İÇİN EĞİTİM SETİ OPTİMİZASYONLU İKİ AŞAMALI KARAR VERME ALGORİTMASI

Öz

   Bu çalışmada, konuşmacı doğrulama görevi için iki aşamalı bir karar verme algoritması önerilmiştir. Bu iki aşamalı algoritma, ikinci aşamada sahtekarlara dayanıklı yapı sayesinde ilk aşamayı geçen sahtekârları ortadan kaldırmayı amaçlıyor. Birinci aşamada, öznitelik olarak mel-frekanslı sepstral katsayılar (MFCC) kullanılarak temel bir sistem oluşturulmuş ve bir radyal taban fonksiyonu (RBF) sinir ağı kullanılarak konuşmacı modellemesi gerçekleştirilmiştir. Ardından, eğitim setini iki kısımda optimize etmek için araştırmalar gerçekleştirildi: (1) taklitçi konuşmacı özniteliklerinin gerçek konuşmacı özniteliklerine oranı, (2) taklitçi konuşmacı kümesi içinde aynı cinsiyet özniteliklerinin zıt cinsiyet özniteliklerine oranı (gerçek konuşmacıya bağlı olarak). Son olarak, iki aşamalı karar verme algoritması sunulmuş ve iki aşamalı sistem tarafından sağlanan performans artışı test sonuçlarıyla birlikte verilmiştir.

Anahtar Kelimeler

• Konuşmacı doğrulama,Eğitim kümesi optimizasyonu,RBF yapay sinir ağları,MFCC,Cohort

Kaynakça

1. [1] WIQAS G., NAVDEEP S., “Literature Review on Automatic Speech Recognition”. International Journal of Computer Applications, 41, 42-50, 2012.
2. [2] SHIKHA G., AMIT P., ACHAL S., “A Study on Speech Recognition System: A Literature Review”, International Journal of Science, Engineering and Technology Research (IJSETR), 3, 2192-2196, 2014.
3. [3] LIU Y, QIAN Y., CHAN N., FU T., ZHANG Y., YU K., “Deep Feature for Text-dependent Speaker verification”, Speech Communication, 73, 1–13, 2015.
4. [4] BHATTACHARYYA S, SRIKANTHAN T, KRISHNAMURTHY P, “Ideal GMM parameters and posterior log likelihood in speaker verification”, Proc. IEEE Signal Processing Soc. Neural Networks for Signal Processing XI, 471-480, 2001.
5. [5] XU Y., SHEN F., ZHAO J., “An incremental learning vector quantization algorithm for pattern classification”. Neural Computing and Applications, 21, 1205–1215, 2012.
6. [6] GALES M., YOUNG S., “The Application of Hidden Markov Models in Speech Recognition”. Foundations and Trends in Signal Processing, 1, 195–304, 2007.
7. [7] PATEL I., SRINIVAS Y. R., “A Frequency Spectral Feature Modeling for Hidden Markov Model Based Automated Speech Recognition” Recent Trends in Networks and Communications, Communications in Computer and Information Science, 90, 134-143. Springer, Berlin, Heidelberg, 2010.
8. [8] KAMRUZZAMAN S. M., A. N. M. REZAUL KARIM A. N. M., ISLAM S., HAQUE E., “Speaker Identification using MFCC-Domain Support Vector Machine”, International Journal of Electrical and Power Engineering, 1, 274-278, 2007.

9. [9] NAIR P. G., NAIR R., “Efficient Speaker Identification Using Artificial Neural Network”, International Journal of Electronics & Communication Technology (IJECT), 6, 27-30, 2015.
10. [10] SWAMY S., SHALINI T., NAGABHUSHAN S.P., NAWAZ S., RAMAKRISHNAN K.V., “Text Dependent Speaker Identification and Speech Recognition Using Artificial Neural Network” Global Trends in Computing and Communication Systems. Communications in Computer and Information Science, 269, 160-168. Springer, Berlin, Heidelberg, 2012.
11. [11] YUE X, YE D, ZHENG C, WU X, “Neural networks for improved text-independent speaker identification”, IEEE Engineering in Medicine and Biology Magazine, 53-58, 2002.
12. [12] MUSTA E., KOMINI V., “A Comparative Study Of Linear Predictive Analysis Methods With Application To Speaker Identification Over a scripting programing”, Journal of Multidisciplinary Engineering Science and Technology (JMEST), 2, 2881-2885, 2015.
13. [13] SINGH A. K., SINGH R, DWIVEDI A., “Mel Frequency Cepstral Coefficients Based Text Independent Automatic Speaker Recognition Using Matlab”, International Conference on Reliability, Optimization and Information Technology (ICROIT), 524-527, Haryana, India, 2014.
14. [14] DAS A., JENA M.R., BARIK K. K., “Mel-Frequency Cepstral Coefficient (MFCC) - a Novel Method for Speaker Recognition”, Digital Technologies, 1,1-3, 2014.
15. [15] PANDIARAJ S., SHANKAR KUMAR K. R., “Speaker Identification Using Discrete Wavelet Transform”, Journal of Computer Science, 11, 53-56, 2015.
16. [16] HALDAR R., MISHRA P. K., “Multilingual Speech Recognition Using Radial Basis Function (RBF) Neural Network”, International Research Journal of Engineering and Technology (IRJET), 3, 2856-2862, 2016 .
17. [17] SHARMA S., SHUKLA A., MISHRA P., “Speech and Language Recognition using MFCC and DELTA-MFCC”, International Journal of Engineering Trends and Technology (IJETT), 12, 449-452, 2014.
18. [18] http://www.phon.ox.ac.uk/files/apps/old_IViE/download1.php (erişim tarihi 08.01.2018)
19. [19] HAYKIN S., Neural Networks a Comprehensive Foundation (2nd ed), Prentice Hall Inc. USA, 1999.
20. [20] JHANWAR N., RAINA., “Pitch Correlogram Clustering for Fast Speaker Identification”, EURASIP Journal on Applied Signal Processing, 17, 2640-2649, 2004.
21. [21] DJEMILI R., BOUROUBA H., KORBA M.C.A., “A Speech Signal Based Gender Identification System Using Four Classifiers”, 2012 International Conference on Multimedia Computing and Systems, 1-4, Tangier, Morocco, 10-12 May 2012.
22. [22] DJEMİLİ R., BOUROUBA H., KORBA M.C.A., O’SAUGHNESSY D., “Boosting Speaker Identification Performance Using a Frame Level Based Algorithm”, International Conference on Communications, Signal Processing, and their Applications (ICCSPA'15), 1-6, Sharjah, United Arab Emirates, 17-19 Feb. 2015.