Chen Xueqin

Performance analysis of mandarin whispered speech recognition based on normal speech training model

The magnitude of the decline in performance is very alarming when the features commonly used in normal speech recognition system are directly used as the input feature of whispered speech in the speech recognition system trained by normal speech. In this paper, in order to finding the characteristics of better matching degree between normal and whispered speech, we propose a spectrum sparse-based approach to obtain the feature of speech spectrum structure. We construct a hidden Markov model based speech recognition baseline system to compare the performance of different features. Experimental results show that the proposed feature can perform better on whispered speech recognition based on the speech recognition system trained by normal speech. This means that the recommended feature is better able to express the similarity between the normal speech and the whisper in the spectral topology structure.

Pitch detection of nosy speech signal based on Bark wavelet transform

On the basis of psychological acoustic theories and experiments, this paper introduces the concept of Bark wavelet, and applies it to pitch detection of noisy speech signals. Experimental results obtained indicate this method is more suitable for noisy speech signals than the classical pure autocorrelation method.

Performance improvement of Mandarin digital whispered speech recognition based on multistage classification

This paper uses digital speech as research object. A baseline system of Mandarin whispered digital speech recognition using Hidden Markov Model is built. In this paper, the performance of the baseline recognition system is analyzed in detail and we find there are three pairs of easily confused speech. Furthermore, the cause of confusion is analyzed in depth. Then a classifier for distinguishing between the three pairs of easily confused digital speech is built to improve the performance. Experiments show that the recognition rate of Mandarin digital whispered speech has been improved greatly, which can be raised from 68.5% to 83.6%.