基于GMM的说话人分类自适应算法

A Speaker Clustering Adaptive Algorithm Based on Gaussian Mixture Model

本文提出了一种基于GMM(GaussianMixtureModel)的说话人分类算法 ,主要介绍了GMM模型和具体实现以及基于GMM的说话人分类算法在关键词检测系统中的应用。实验结果证明 ,该算法计算量小、实现速度快 ,能够提高系统的识别率 ,扩大应用人群范围 。

In this paper,a speaker clustering adaptive algorithm based on the Gaussian Mixture Model is provided The Gaussian Mixture Model is described and how to get the GMM parameters is also introduced The speaker clustering adaptive algorithm is applied to the keyword spotting systems The experimental result shows that an average detection rate improvement of 3 2% can be obtained by using the proposed adaptive algorithm with small computation and fast implementation This algorithm can be expected to be applied to a larger group of people and shows a wide prospect of application in small to intermediate range vocabulary recognition systems