A new noisy speech recognition methodi
Xiaoqun Zhao, Jinmin Wang · 2006
The use of speech recognition techniques in many practical applications has demonstrated the need for improved algorithm to suppress the effect of the background noise. This paper describes a new speech recognition method based on morphology that provides robust performance in noisy environments. Morphological operations are nonlinear signal transformations that locally modify geometric features of signals. The morphological filter is firstly used as the front-end noise suppression algorithm for noisy speech data that will be tested by a speech recognition system. But on the other hand the enhanced speech has some distortions compared to the original clean speech, causing loss of some speech details and errors in recognition. Therefore, the same morphological filter is used to preprocess the training speech examples before the training phase. The experimental results show that the performance of speech recognition systems can be improved effectively by using the new method under noisy environment.