摘要
在声纹识别系统中,由于特征参数MFCC不能实现高效的识别,提出了将MFCC和共振峰相结合的提取方法。并在此基础上引用了一阶差分ΔMFCC和二阶差分ΔΔMFCC,进而将MFCC、ΔMFCC、ΔΔMFCC和共振峰有机结合起来进行声纹识别实验,说话人模型采用的是高斯混合模型。实验结果表明提取混合特征参数MFCC、ΔMFCC、ΔΔMFCC与共振峰进行声纹识别时,识别率大大提高。
Because MFCC can＇t achieve efficient voiceprint recognition, a feature extraction method by combining MFCC and Formants is proposed. And on the basis cited the A MFCC and AA MFCC, then MFCC, A MFCC, AA MFCC and Formants are combined to conduct the voiceprint recognition experiment, the speaker models use the Gaussian Mixture Modeling. The ex- periment results show that the extraction of mixture parameters MFCC, A MFCC, AA MFCC and Formants for the voiceprint recognition, the recognition rate improves greatly.
作者
王正创
WANG Zheng-ehuang （College of Mechanical and Electronic Engineering, Chaohu University, Chaohu 238000,China）
出处
《电脑知识与技术》
2016年第2期188-190,共3页
Computer Knowledge and Technology
基金
巢湖学院基金项目（XLZ-201505）
