Abstract
When a speech recognition system is used in real environment, noises cause a serious problem in the system. In this paper, we propose a method which normalizes feature parameters after filtering the time trajectories of both the power spectra and the log power spectra. As a result of the isolated word recognition experiment where fifteen kinds of noises were added, higher recognition rates was obtained than conventional method by Spectrum Subtraction (SS) and RelAtive SpecTrAl (RASTA). But the proposed method has problems such as processing time and groupe delay of FIR filter. This paper reports on the outline of LSI to solve these problems.
