AN EFFICIENT SVM BASED SPEAKER VERIFICATION SYSTEM

Author(s):

P.Suganthi , Christian College of Engineering and Technology, Oddanchatram,India; M.N.Srinithya, Christian College of Engineering and Technology, Oddanchatram,India

Keywords:

Gaussian kernel Processing Element(GK-PE), Scaling Unit, Discrete Fourier Transform(DFT), Coordinate Rotate Digital Integrated Circuit(CORDIC), low power

Abstract:

The speaker verification system is to identify information from speaker voice and this system is mainly used to voiceprints and bio metric identification process. In the proposed system, the useful information from speech signal is extracted using DFT methodology. The SVM module will generate test vector for the input speech signal. The SVM module consists of Gaussian Kernel unit and Scaling unit. The Gaussian kernel unit consist of Gaussian kernel processing element which is used to evaluate kernel values for test vector and support vector. The existing system uses CORDIC circuit in exponential processing element to minimize architecture. This work uses several adders and shifters to minimize the shifting and rotation operation. The proposed system uses approximate multiplier circuit in the normal processing element to minimize architecture. The scaling unit performs multiplication and the remaining operations of decision value evaluation. Finally, the spoken words and claimed speaker is verified. The power and circuit complexity is reduced in this system.

Other Details:

Manuscript Id	:	IJSTEV2I10055
Published in	:	Volume : 2, Issue : 10
Publication Date	:	01/05/2016
Page(s)	:	219-224

Download Article