提出了一种隐马尔可夫模型（ＨＭＭ）和径向基函数神经网络（ＲＢＦ）相结合的语音识别新方法。该方法首先利用ＨＭＭ生成最佳语音状态序列，然后用函数逼近技术产生对最佳状态序列进行时间规正，最后通过ＲＢＦ神经网络进行分类识别。理论和实验结果表明，该系统比ＨＭＭ具有更好的识别效果，特别对提高易混淆词的识别性能尤为显著。 A new speech recognition method based on Hidden Markov Model (HMM) and radial basis function neural network (RBF) is proposed. Firstly, HMM is used to generate the best speech state sequence, and then the function approximation technique is used to generate the best time series. Finally, RBF neural network is used for classification and recognition. Theoretical and experimental results show that this system has a better recognition effect than HMM, especially for improving the recognition performance of confusing words.