单通道的语音盲去混响  

Speech Dereverberation of Single Channel

作　　者：沈希忠[1,2] 孟光[2] 

机构地区：[1]上海应用技术学院机械与自动化学院,上海200235 [2]上海交通大学机械系统与振动国家重点实验室,上海200240

出　　处：《上海交通大学学报》2010年第2期229-233,共5页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金重点项目(10732060);上海市教育委员会重点学科建设项目(J51501);上海市教委一般资助项目(050Z06)

摘　　要：针对单通道的语音盲去混响研究中存在的问题,提出了一种盲去混响的方法.首先对语音信号进行可变长度的分段,然后采用复倒谱技术,对房间声学冲激响应进行预估计,得到去混响的初始解,并以此作为盲解卷积算法的初始解进一步去混响.仿真和实验表明,该方法具有很好的去混响效果.A new method of blind dereverberation of single channel was proposed to solve the existing problem in speech signal processing. Firstly, segmentation of speech signal with variant length is carried on the sampled speech signal, and the complex cepstrum is applied to the valid segments. Then, the pre-estima- tion of the room impulse response is obtained by this method, and so is the inverse-filter solution. At last, blind deconvolution is used by taking the inverse-filter as the initial parameter of the algorithm. The simulations and experiments show its validity of the proposed method.

关 键 词：去混响 复倒谱 盲解卷积 语音 分段 

分 类 号：TN911.72[电子电信—通信与信息系统]