A Novel Approach for Speaker Recognition by Vowel Specking

S. Nasir, I. Khokar, and M.N. Jafri (Pakistan)

Keywords

Speech processing, wavelet transform, multiresolution analysis

Abstract

This paper presents a novel approach to speck vowels by using wavelet transform. The wavelet transform not only produces a flexible decomposition of signals in both time and frequency domain, which is useful in segmenting parts of speech containing fast and slow varying changes but is also helpful in highlighting the frequency distribution contrast which exists among different classes of phonemes. These vowels can be used to develop speaker dependent phoneme models for the purpose of speaker recognition.

Important Links:



Go Back