基于加权混合特征的话者识别算法  

Speaker recognition algorithm based on weighted mixture features

作　　者：徐志江[1] 赵梦娜 卢为党[1] 

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310023

出　　处：《浙江工业大学学报》2017年第6期628-633,共6页Journal of Zhejiang University of Technology

基　　金：国家自然科学基金资助项目(61471322;61402416)

摘　　要：用多窗谱估计和伽马通滤波改进经典的梅尔倒谱特征(MFCC)的识别性能,并与delta特征相结合,提出了一种基于加权参数设置策略的混合特征话者识别算法.该算法解决了梅尔倒谱系数方差过大、听觉特征不明显及话者识别算法特征单一的问题.仿真结果表明:与MFCC和线性预测的提取方法相比,该算法鲁棒性能更优,对不同噪声环境的适应性更好.Multi-window spectrum estimation and gamma-pass filtering are used to improve the recognition performance of classical Mel-cepstral feature(MFCC).Combined with the delta feature,a mixed feature speaker recognition algorithm based on weighted parameter setting strategy is proposed.The algorithm solves the problem that the Mel-cepstral coefficient variance is too large,the auditory features are not obvious,and the feather of speaker recognition algorithm is simple.The simulation results show that the proposed algorithm has better robust performance and better adaptability to different noise environments than MFCC and linear predictive extraction methods.

关 键 词：多窗谱估计 伽马通滤波器组 加权函数 加权混合特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]