摘要
引入两种新的特征参数Mid-MFCC和IMFCC,采用MFCC、Mid-MFCC和IMFCC相结合的改进算法,解决MFCC特征参数在语音识别中对中、高频信号的识别精度不高的特点,并使用增减分量法计算MFCC、Mid-MFCC和IMFCC各阶倒谱分量对语音情感识别的贡献,提取3个特征参数贡献最高的几阶倒谱分量组成了新的特征参数;实验结果表明,在相同环境下新的特征参数比经典MFCC特征参数的语音情感的识别率稍高。
This paper introduced two new characteristic parameters Mid-- MFCC and IMFCC combining with MFCC to improve the algo- rithm which solve the problem that MFCC characteristic parameter in speech recognition has low identification accuracy when signal is inter- mediate, high frequency signal, calculating the contribution that MFCC, Mid--MFCC and !MFCC each order cepstrum component was used in speech emotion recognition with increase or decrease component method, extracting highest contribution of several number order cepstrum component from three characteristic parameters and forming a new characteristic parameter. The experiment results show that new character- istic parameter has higher recognition rate than classic MFCC characteristic parameter in speech emotion recognition under the same environ- ment.
出处
《计算机测量与控制》
北大核心
2013年第7期1966-1968,1986,共4页
Computer Measurement &Control
基金
国家自然科学基金资助项目(60961002)
广西自然科学基金资助项目(2012GXNSFAA053221)
