基于模糊聚类的矢量量化的声纹识别研究

Voice Recognition Based on Fuzzy Clustering and Vector-Quantization Algorithm

采用模糊C-均值聚类算法(FCM)实现声纹码本的矢量量化,使用基于相似系数和的孤立点检测法识别孤立点.试验表明,该方法能有效地减少孤立点对识别结果的干扰,显著降低码本量化误差,从而提高矢量量化声纹识别系统的识别率.

In this paper,a fuzzy c-means（FCM） based Vector Quantization（VQ） approach is applied to the codebook of a voice recognition system,and then the isolated point detection method based on the sum of similarity coefficients is used to find out those isolated points.The experimental results show that the isolated point detection method can effectively exclude the interference of the outliers on the clustering results,and that the FCM based VQ technique reduces the quantization error remarkably.The recognition rate of VQ based voice recognition system can be improved by the proposed approach.