Abstract: In order to overcome the inadequate of emotional conventional linear argument at depicting different types of character sentiments, this paper takes the multiple fractals theory into the sound emotional identify, by analyzing the multiple fractal features on the different sound emotional state, and proposes multifractal spectrum parameters and generalizes hurst index as emotional conventional parameters, combines with traditional voice acoustic features and using Support Vector Machine（SVM） for speech emotion recognition. The results show that the accuracy and stability of the recognition system are improved effectively through using non-linear parameters, compared with the linear features of traditional voice recognition method, so it provides a new idea for voice emotion recognition.

Key words: multifractal, Hurst exponents, emotion speech, emotion recognition

摘要： 为了克服语音情感线性参数在刻画不同情感类型特征上的不足，将多重分形理论引入语音情感识别中来，通过分析不同语音情感状态下的多重分形特征，提取多重分形谱参数和广义Hurst指数作为新的语音情感特征参数，并结合传统语音声学特征采用支持向量机（SVM）进行语音情感识别。实验结果表明，通过非线性参数的介入，与仅使用传统语音线性特征的识别方法相比，识别系统的准确率和稳定性得到有效提高，因此为语音情感识别提供了一个新的思路。

关键词: 多重分形, Hurst指数, 语音情感, 情感识别