一种基于脉冲耦合神经网络的语音情感识别新方法

Innovative method for emotion recognition of speech based on pulse coupled neural network

针对汉语语音情感识别问题,提出了一种基于脉冲耦合神经网络(PCNN)的识别方法。该方法将语音转化为语谱图后输入到PCNN,得到输出图像的神经元点火序列及其熵序列作为语音情感的特征,利用其特征实现语音情感识别。实验结果表明,该方法可以有效地识别"高兴"与"平常"这两种不同的情感。该方法将PCNN引入到语音情感识别的应用研究中,开拓了语音和图像信号结合处理的新领域,同时对于PCNN的理论研究和实际应用具有重要的现实意义。

An innovative method for classifying emotional states of neutral or happy of Chinese phonetics by using Pulse Coupled Neural Network （PCNN） was proposed. The entropy series and neurons firing series of the image obtained by feeding the spectrogram into PCNN were used as the characteristics of emotion speech for emotion recognition, The experimental result show that this method can distinguish speaker＇s emotions（ normal ＆ happy） in speech effectively. The application of PCNN in emotion recognition of speech expands the combination of two important parts of signal processing-speech and image processing and it is significant for theoretical research and application of PCNN.