Authors: Engin ERZİN, Yasemin YARDIMCI, A. Enis ÇETİN
Abstract: In this paper, a new set of feature parameters for speech recognition is presented. The new feature set was obtained from the root-cepstral coefficients derived from the wavelet analysis of the speech signal. The performance of the new feature representation was compared to mel scale cepstral representation (MELCEP) in the presence of car noise. Subband analysis based parameters were observed to be more robust than the commonly employed MELCEP representation.