Abstract: The paper proposes a method to recognize speech by Improved Pulsed-Coupled Neural Network(IPCNN). IPCNN is first used to extract the feature of spectrogram，and is then with the assistance of Probabilistic Neural Network(PNN) to recognize speech. By training speech samples，speech identification databases are constructed，and the integrated recognition system is then built. The experiment results show that comparing with using the IPCNN and PNN alone,this method can increase the recognition rate by 22.7% and 39.4% .And it can attain recognition rate of 92%.

Key words: Probabilistic Neural Network, Spectrogram, Speech recognition, Pulse-Coupled Neural Network

摘要： 提出了一种改进脉冲耦合神经网络(IPCNN)实现语音识别的方法。首先利用IPCNN来快速提取语音的语谱图图像特征，然后由概率神经网络(PNN)辅助来识别语音。通过训练语音样本来构成语音识别库并建立综合识别系统。实验结果表明,本方法相对于单独使用PCNN和PNN识别率分别提高了22.7%和39.4%,达到92%的识别率。

关键词: 概率神经网络, 语谱图, 语音识别, 脉冲耦合神经网络