Subband-Autocorrelation analysis and its application for speech recognition

In this paper, the Subband-Autocorrelation (SBCOR) analysis technique is proposed, and applied to speech recognition. Although the SBCOR analysis is a simplified version of the auditory model proposed by Seneff, it is not an auditory model, but a signal analysis technique based on filter bank and autocorrelation analysis. The SBCOR analysis system is evaluated for five types of filter banks and three autocorrelation detectors using a speaker dependent DTW word recognition system. The experimental results show that the SBCOR spectrum performs as well as the smoothed group delay spectrum under clean condition, and much better under noisy conditions. Finally, it is shown that the optimum filter bank is a fixed Q filter bank whose center frequencies are equally spaced on the Bark scale, and the optimum autocorrelation analysis is a conventional autocorrelation detection without controlling weak signals. An analysis example of the SBCOR is also shown