Research on emotional speech recognition based on pitch
基于基频特征的情感语音识别研究

This paper first analyzed the change of pitch in speech signal under four emotions such as angry, happy, neutral and sad. Then twelve pitch related emotional features were extracted and used to train the Gaussian mixture models (GMMs) for each emotional state. In emotion recognition, the emotional state with the GMM having the maximal likelihood ratio was determined as the final out. The experimental result shows that the pitch related features and GMM are effective in emotion recognition from speech.