摘要
随着社会和科技的快速发展,如何有效识别语音情感已经成为人们关注的一个热点。在众多的分类算法中,Adaboost多分类算法得到了较好的应用效果。该算法将生气、开心、中性、伤心和害怕5种语音情感分为三层,由粗到细,逐层识别。基于柏林情感语音库,将提取的语音情感特征利用Fisher准则选择较佳特征作为实验数据,实验结果表明,将相近情感分在一起训练更有利于提升Adaboost算法的分类性能。此外,在与传统的BP和SVM分类模型比较中,Adaboost多分类算法表现出了优越性。
With the rapid development of society and technology,how to effectively identify speech emotion has become a hot topic. In many classification algorithms,Adaboost multi-classification algorithm has a good application effect. The algorithm divides 5 speech emotions （anger,happiness,neutral,sadness and fear）into three layers. It is identified layer by layer from coarse to fine.Based on Berlin emotional speech database,the speech emotion features are extracted,and the better features are selected as the experimental data using Fisher criterion. The experimental results show that it is more advantageous to improve the classification performance of Adaboost algorithm by training similar emotions. In addition,Adaboost multi-classification algorithm shows superiority in comparison with traditional BP and SVM classification models.
作者
邢尹
刘立龙
程胜
时满星
李月锋
XING Yin;LIU Lilong;CHENG Sheng;SHI Manxing;LI Yuefeng(School of Geomatics and Geoinformation,Guilin University of Technology,Guilin 541004;School of Civil Engineering,Lanzhou University of Technology,Lanzhou 730050)
出处
《计算机与数字工程》
2018年第11期2197-2200,2229,共5页
Computer & Digital Engineering
