Comparison of Two Speaker Recognition Systems
A.Vaishnavi1, B. Chanakya Raju2, G.Prathiksha3, L. Harshitha Reddy4, C. Santhosh Kumar5
1A.Vaishnavi, Electronics and Communication, Amrita Vishwa Vidyapeetham, Coimbatore, India.
2B.Chanakya Raju, Electronics and Communication, Amrita Vishwa Vidyapeetham, Coimbatore, India.
3G.Prathiksha, Electronics and Communication, Amrita Vishwa Vidyapeetham, Coimbatore, India.
4L.Harshitha Reddy, Electronics and Communication, Amrita Vishwa Vidyapeetham, Coimbatore, India.
5C.Santosh Kumar, Electronics and Communication, Amrita Vishwa Vidyapeetham, Coimbatore, India.
Manuscript received on March 27, 2014. | Revised Manuscript received on April 05, 2014. | Manuscript published on April 30, 2014. | PP: 300-303 | Volume-3, Issue-4, April 2014. | Retrieval Number: D2997043414/2013©BEIESP
Open Access | Ethics and Policies | Cite
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: This paper presents a comparison between two speaker recognition systems. One system uses 30 Shannon entropy values extracted from a four level wavelet packet decomposition method in addition to the first three formant frequencies as features and a cascaded feed forward back propagation neural network is used as classifier. The second system uses Mel frequency cepstral coefficients (MFCC) as features and a support vector machine (SVM) as classifier. Results suggest that wavelet based system has better performance than the classic MFCCs with an efficiency of 89.56%.
Keywords: Shannon entropy, Formant frequencies, cascaded neural network, MFCC, SVM.