基于AR模型的Kalman语音增强算法  

Kalman Speech Enhancement Algorithm Based on AR Model

作　　者：王涛[1] 鲁怀伟[1] 刘宝成 WANG Tao;LU Huai-wei;LIU Bao-cheng(School of Mathematics and Physics,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学数理学院,兰州730070

出　　处：《青岛大学学报（自然科学版）》2018年第2期48-53,共6页Journal of Qingdao University(Natural Science Edition)

摘　　要：针对AR模型参数估计的精度不高和无法解决谐波噪声等问题,结合Kalman滤波对语音增强算法进行改进,提出了基于AR模型的Kalman语音增强算法,AR模型的Kalman语音增强算法主要对数据进行预处理,剔除奇异点,然后建立AR模型,求出模型的输出的卡尔曼滤波。实验结果表明,AR模型的参数估计精度得到了较大提高,语音增强效果较好。According to the AR model parameter estimation accuracy is not high and cannot solve the problem of harmonic noise,speech enhancement algorithm combined with Kalman filter,Kalman speech enhancement algorithm is proposed based on AR model,AR model Kalman speech enhancement algorithm of data preprocessing,eliminating singularity point,then set up AR model,obtained by Calman the output of the filter model.The experimental results show that the precision of parameter estimation of AR model is greatly improved and the effect of speech enhancement is better.

关 键 词：AR模型 KALMAN滤波 语音增强 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]