过零周期转移概率矩阵语音识别方法的改进

An Improued Speech Recognition Method on the Basis of the Algorithm of the Zero-cross Transfer Probability Matrix.

本文在过零周期序列转移概率算法的基础上，提出了一种改进方法，即从语音的过零周期序列中，找出其“过渡音段”，给以加权。通过对特定人孤立字（０～９十个数字）的实验，结果表明改进后的方法识别率有所提高，但并未使算法复杂化。

Based on the algorithm of the zezo cross transfer probability matrix, this paper in-troduces an improved speech recognition method, which involves pick out transitionspeech segments and having them weighted. The experiment of the dependent speaker andisolated words recognition shows that the recognition rate has been increased without com-plicating the algorithm.