基于改进的Kalm.pdfVIP

第 28 卷第 6 期 声 学 技 术 Vol.28, No.6 2009 年 12 月 Technical Acoustics Dec., 2009 基于改进的Kalman 滤波的语音增强算法 1,2 2 2 2 余 华 ，陈国明 ，赵 力 ，邹采荣 (1. 南京信息职业技术学院，南京 210023 ；2. 东南大学信息科学与工程学院，南京 210096) 摘要：传统的 kalman 滤波方法在推导过程中假定观测噪声为白噪声。通常对于有色噪声需要用白噪声激励的方法予 以模拟，并且需要以牺牲运算量作为代价。本文提出了一种改进的基于kalman 滤波的语音增强算法，可以处理白噪 声和有色噪声情况，不需要增加计算量，仿真结果表明了该算法对有色噪声的语音增强性能要优于基于传统 kalman 滤波方法。 关键词：语音增强；卡尔曼滤波；有色噪声 中图分类号：TN912.35 文献标识码：A 文章编号：1000-3630(2009)-06-0763-05 DOI 编码：10.3969/j.issn1000-3630.2009.06.016 Speech enhancement based on improved Kalman filtering method 1,2 2 2 2 YU Hua , CHEN Guo-ming , ZHAO Li , ZOU Cai-rong (1. Nanjing Information Occupation Technical College , Nanjing 210023, China; 2. School of Information Science and Engineering, Southeast University, Nanjing 210096, China) Abstract: In this paper a single channel speech enhancement method for noisy speech signals is proposed, which is based on improved Kalman filter. Due to the high computation cost and the need of accurate estimation of AR model of speech in colored noise, a subspace based speech enhancement method is introduced. The proposed method can make tradeoff between residual noise and speech distortion. Experimental results demonstrate the improved performance of the proposed algorithms. Key words: speech enhancement; Kalman filter; colored noise 1 引 言 应用。Gibson[3]提出了一种次最优的 Kalman 滤波方