论文部分内容阅读
变帧率（ＶＦＲ）分析技术可以在语音识别时，用以丢弃那些特征非常相似的语音帧．文中分析了已有ＶＦＲ方法的不足，并提出了一种新的ＶＦＲ方法．这种新方法能够更加突出语音信号中发音变化区域．计算机模拟实验显示，经该方法预处理后的语音送入隐马尔柯夫模型的语音识别系统比传统算法有更高的识别率．
Variable frame rate (VFR) analysis techniques can be used to discard speech frames that are very similar in speech recognition. In this paper, we analyze the shortcomings of the existing VFR methods and propose a new VFR method. This new method can give more prominence to the change region of speech in speech signals. Computer simulations show that the speech recognition system using the preprocessed speech into the hidden Markov model has a higher recognition rate than the traditional algorithm.