Study on the Chinese continuous speech recognition under noise environments based on PCANN/HMM

摘要

This paper presents a method to improve the noise robustness of Chinese continuous speech recognition system based on the PCANN/HMM hybrid structure. By using the principal components combined by successive multi-frames as the input of HMM, it introduces the dependency between frames and also reduces the noise effectively. And in this paper, we also improve the traditional spectral subtraction method. Experimental results demonstrate the efficiency of the new algorithms in Chinese continuous speech recognition under high noisy environments.