摘要
针对在语音情感识别中孤立使用隐马尔科夫模型(HMM)固有的分类特性较差的缺点,本文提出了利用隐马尔科夫模型和径向基函数神经网络(RBF)对惊奇,愤怒,喜悦,悲伤,厌恶5种语音情感进行识别的方法。该方法借助HMM规整语音情感特征向量,并用RBF作为最终的决策分类器。实验结果表明在本文的实验条件下此方法和孤立HMM相比具有更好的性能,厌恶的识别率有了较大该进。
The paper have presented a method for human speech emotion recognition which includes surprise, anger, joy, sadness and disgust by combining Hidden Markov Model and Radial Basis Function neural network towards the poor class discriminate deficiency in traditional isolated HMM, The approach employs HMM to arrange speech emotion feature vector and RBF as the final decision stage classifier, The result shows that the method has better performance than isolated HMM with the recognition rate of disgust emotion improving a lot,
出处
《微计算机信息》
北大核心
2007年第34期218-219,296,共3页
Control & Automation
关键词
隐马尔科夫模型
径向基函数神经网络
空间正交基
语音情感识别
Hidden Markov Model, Radial Basis Function Neural Network, Orthogonal Basis, Speech Emotion Recognition
