基于模糊核Fisher判别的说话人识别

speaker recognition based on fuzzy kernel Fisher discriminant

在深入研究核Fisher判别方法的基础上,提出一种新的模糊核Fisher判别算法应用于说话人识别。采用模糊C均值聚类方法选择样本数据的同时,得到样本的模糊隶属度矩阵和聚类中心向量,进而对核Fisher判别算法中的类间离散度矩阵和类内离散度矩阵进行改进,生成模糊核Fisher判别算法,将其应用于说话人语音识别。

Based on the in-depth study of kernel Fisher discriminant, a novle speaker recognition approach based on Fuzzy Kernel Fisher Discriminant was proposed in this paper. The training data was selected by using fuzzy C-means clustering, simultaneously the class center matrix and the fuzzy membership matrix can be achieved to redefine between-class scatter matrix and within-class scatter matrix of kemel fisher discriminant. Then a novel fuzzy kernel fisher discriminant was proposed to apply in speaker recognition.