Auditory Filterbank Based Perceptual Linear Prediction for Speech Recognition in Noisy Environment

Modern automatic speech recognition (ASR) systems typically use a bank of linear filters as the first step in performing frequency analysis of speech. On the other hand, the cochlea, this is responsible for frequency analysis in the human auditory system. It will be shown in this paper that it presents a new design to extract the feature based on the human auditory system characteristics, it relies on the gammachirp filter bank. The prosodic features such as jitter and shimmer are extracted from the fundamental frequency contour and added to baseline spectral features, specifically, Perceptual Linear Prediction (PLP) for human speech, Gammachirp Filterbank Perceptual Linear Prediction (GFPLP) and Modified Gammachirp Linear Prediction (MODGLP). The results show that the new model with gammachirp gives results that are comparable to ones obtained by PLP and GFPLP. Experimental results show the best performance of this architecture. The above-mentioned techniques were tested with impulsive signals under various noisy conditions within AURORA databases.