用DSP实现的一个简单的语音识别系统,只要实现单个词的识别即可,采样率8k,帧长30ms,帧移10ms,系统采样后分帧--端点检测,将检测到的原始语音信号保存下来,基本上一个字在30帧左右,然后提取每帧的LPC参数--将LPC参数转换为LPC倒谱系数,然后利用DTW方法和模板比较.-a brief speech recognition system, as long as the realization of a single word can be identified. 8 k sampling rate, fr a me size of 30 ms and 10 ms fr a me shift, hours after sampling fr a me-- Endpoint Detection, Detection of the original speech signal preserved, basically a word in 30% Then each fr a me from the LPC parameters-- will be converted to LPC parameters LPC cepstrum coefficients. Then DTW methods and template comparison.