该文设计了以MCU+DSP的双CPU结构为核心的语音识别系统,并针对语音识别系统的预处理和特征提取阶段,引入小波变换对传统的实现方法进行改进,有效地降低了系统中的噪声,模式识别中采用了隐马尔可夫模型,提高了系统的语音识别性能.实验数据验证了该方法的可行性和有效性.“,”A speech recognition system based on double CPU cores MCU+DSPis designed in the paper. Wavelet transform methods are introduced into preprocess and characteristics extraction stages of the system for improvements of traditional methods, which decrease noises of the system effectively. Hidden Markov model is adopted in the pattern recognition, which improves the performance of the speech recognition greatly. The results of experiments shows the validity and the feasibility of the method proposed.