基于自适应小波收缩的抗噪声说话人识别  

Anti-noise speaker recognition based on adaptive wavelet shrinkage

作　　者：姜娇娇[1] 罗飞[1] 杨淑莹[1] JIANG Jiao-jiao LUO Fei YANG Shu-ying(School of Computer and Communication Engineering, Tianjin University of Technology, Tianjin 300384, Chin)

机构地区：[1]天津理工大学计算机与通信工程学院,天津300384

出　　处：《天津理工大学学报》2016年第6期5-8,19,共5页Journal of Tianjin University of Technology

摘　　要：说话人识别面临许多实际困难,其中由于环境和采集通道因素导致信号不一致最具挑战性.在本文中,提出一种新的自适应小波收缩的抗噪声说话人识别方法.在小波收缩去噪的应用中,双阈值策略压缩抑制噪声,保留信号系数,用重叠语音信号端的梅尔倒谱系数的修正来识别.用两个公共可用的语音信号数据库来评价所提出的方法的有效性,并与其它方法相比.证明了所提出的方法在不同的噪声条件下具有更好的鲁棒性.Speaker recognition faces many practical difficuhys, the inconsistency between environmental and acquisition channel is most challenging. In the paper, a new anti-noise speaker recognition method based on adaptive wavelet shrinkage is proposed. In the application of wavelet shrinkage for noise removal, a dual-threshold strategy is developed to suppression noise and preserve signal coefficients. The recognition is achieved using modification of Mel-frequency cepstral coefficient of overlapped voice signal segments. The two public available speech signal databases are used for evaluate the efficacy of the proposed method. The efficacy of the proposed method is compared with others methods. It is show that the proposed method have better robustness in various noise conditions.

关 键 词：说话人识别 噪声抑制 小波 特征提取 

分 类 号：TN912[电子电信—通信与信息系统]