基于HMM和PNN的语音情感识别研究

Emotion Recognition in Speech Based on HMM and PNN

语音情感识别是从语音的角度赋予计算机理解情感特征的能力,最终使计算机能像人一样进行自然、亲切和生动的交互。提出了一种融合隐马尔科夫模型(hidden mark-ov model,HMM)和概率神经网络(probabilistic neural network,PNN)的语音情感识别方法。在所设计情感识别系统中,提取出基本的韵律参数和频谱参数,利用PNN处理声学参数的统计特征,利用HMM处理声学参数的时序特征,运用加法规则和乘法规则融合了统计特征和时序特征的识别结果。实验结果显示,所提出的算法在语音情感识别中具有有效的识别能力。

The aim of the emotion recognition is make the computer have the capacity of understand emotion by the way of voice characteristics studies and ultimately like people for natural, warm and lively interaction. A speech emotion recognition algorithm based on HMM （hidden Markov model） and PNN （probabilistic neural network） was developed, in the system, the basic prosody parameters and spectral parameters were extracted first, and then the PNN was used to model the statistic features and HMM to model the temporal features. The sum and product rules were used to combine the probabilities from each group of features for the final decision. Experimental results approved the capacity and the efficiency of the proposed method.