电子器件2017，Vol.40Issue(3)：545-550,6.DOI:10.3969/j.issn.1005-9490.2017.03.006
基于GMM非线性变换的说话人识别算法的研究
Text-Independent Speaker Recognition Using GMM Non-Linear Transformation
摘要
Abstract
For the text independent speaker recognition GMM model,some non-target models of the test frame of the model score may be relatively high,thus causing the problem of false.Based on the statistical properties of the frame likelihood probability,a GMM nonlinear transformation method is proposed.This method gives different weights to each frame model,which makes the model with high score and low weights,as the target model score higher than other non target frame model,so it can improve the total score of the target model,reduce the score of non target model,thus reducing the possibility of false positives.Theoretical results and experimental results show that the proposed method can improve the recognition rate of GMM speaker recognition.关键词
与文本无关说话人识别/混合高斯模型/非线性变换Key words
text-independent speaker recognition/Gaussian mixture model/non-linear transformation分类
信息技术与安全科学引用本文复制引用
罗文华,杨彦,齐健,赵力..基于GMM非线性变换的说话人识别算法的研究[J].电子器件,2017,40(3):545-550,6.基金项目
国家自然科学基金项目(61301219) （61301219）
2014年青蓝工程资助项目 （）
2015年农业科技创新专项引导资金项目 （）
2015年盐城市农业科技指导性项目(YKN2015031) （YKN2015031）
