Robust Speech Recognition Using Wavelet Domain Front End and Hidden Markov Models

This paper presents a method to address the issue of noise robustness using wavelet domain in the front end of an automatic speech recognition (ASR) system, which combines speech enhancement and the feature extraction. The proposed method includes a time-adapted hybrid wavelet domain speech enhancement using Teager energy operators (TEO) and dynamic perceptual wavelet packet (PWP) features applied to a hidden Markov model (HMM)–based classifier. The experiments are performed using the HTK toolkit for speaker-independent database which are trained in a clean environment and later tested in the presence of AWGN. It has been seen from the experimental results that the proposed method has a better recognition rate than the most popular MFCC-based feature vectors and HMM-based ASR in noisy environment.