基于压缩感知的语音压缩与重构  

Compression and Reconstruction of Speech Signal Based on Compressed Sensing

作　　者：叶蕾[1] 杨震[2] 

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003 [2]南京邮电大学信号处理与传输研究院,江苏南京210003

出　　处：《南京邮电大学学报（自然科学版）》2010年第4期57-60,共4页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(60971129)资助项目

摘　　要：基于语音信号在离散余弦基下的近似稀疏性,对语音信号采用压缩感知技术进行压缩和重构,研究了分析窗长固定时重构误差与观测点数的关系,针对低压缩比下重构信号"noisy"的特性,提出对重构信号进行小波去噪联合低通滤波的方法以改善重构语音的质量,并研究了低压缩比下分析窗长对重构信号质量的影响。仿真结果表明低压缩比下,合理选取分析窗长,并采用小波去噪联合低通滤波的处理方法可以明显改善重构语音的质量。Based on the approximate sparsity of Speech Signal in Discrete Cosine basis,compressed sensing theory is applied to compress and reconstruct speech signal. The relationship between reconstruction error and numbers of measurement is studied under fixed length of analysis window. Aiming at the ＂nois- y＂ characteristic of reconstruction signal with low compression rate, wavelet denoising associated with low pass filter is proposed to improve the quality of CS reconstruction signal. Also the influence of length of analysis window on quality of CS reconstruction signal under low compression rate is investigated. The sim- ulation results indicate the proper selection of window length and method combining wavelet denoising with low pass filter can improve the quality of CS reconstruction signal substantially under low compression rate.

关 键 词：压缩感知 离散余弦变换 线性规划 小波去噪 

分 类 号：TN912.3[电子电信—通信与信息系统]