A NEW FEATURE EXTRCTION MOTIVATED BY HUMAN EAR

Feature extraction method is a very important factor that has great effect on the performance of speech recognition systems. Since living creatures have usually been the best models for human technology, we hope these models can overcome the problem of mismatch between training and testing conditions. In this paper we present a novel feature extraction based on human ear physiology. In this new method, with inspiration of human ear physiology, we model the inner ear for feature extraction. The proposed method is evaluated in a connected digit recognition task. Experimental results show that the proposed feature extraction, called Human Ear based Feature Extraction (HEFE), decreases the relative word error rate by more than 32 % with respect to MFCC features. 1