摘要
在利用机器学习方法进行语音情感识别时,会采用大量的特征,这些特征的冗余降低了识别准确率,加大了计算量和建模时间。利用互相关特征选择(Correlation-Based Feature Selection,CFS)方法对遗传算法(Genetic Algorithm,GA)进行改进,对输入特征进行降维,可使原有的机器学习算法快速收敛,提高了识别正确率。在此基础上设计基于Lab VIEW的语音情感识别系统,实验证明:该系统可以对语音信号进行有效的情感识别。
At the time of making use of machine learning method to recognize emotional speeches, a large number of features including redundancy characteristics which reduces the recognition accuracy and increase the calculation and modeling time were adopted. In this paper, an improved GA algorithm which combining with CFS algorithm was proposed to reduce the input feature dimensions and to rapidly converge the original machine learning algorithm as well as to improve accuracy of identification. Based on this, the LabVIEW- based emotional speech recognition system was designed. Experimental results show that, this system can rec- ognize the signals of emotional speeches effectively.
出处
《化工自动化及仪表》
CAS
2018年第3期205-211,共7页
Control and Instruments in Chemical Industry
关键词
语音情感识别
机器学习
遗传算法
互相关特征算法
识别正确率
emotional speech recognition, machine learning, genetic algorithm,CFS algorithm, recognition accuracy
