摘要
根据语音信号经过小波分解后低频分量和高频分量的特点,提出分别对他们进行自适应压缩感知。首先对信号的低频分量用训练的过完备基进行稀疏分解,降低了稀疏分解过程中的计算量。然后详细描述了改进自适应观测矩阵的产生,以及对低频和高频分量分别进行自适应观测。最后通过OMP重构算法分别对低频和高频分量进行重构,通过小波合成还原出原始信号。实验表明,语音信号在基于小波分解的自适应压缩感知方案中具有良好的重构性能。
Based on the characteristics coefficients of speech signal at low frequency and high frequency after wavelet transformation,this paper proposes adaptive speech compressed sensing.First,the trained overcomplete dictionary is applied to the low frequency coefficients after wavelet transformation to decrease the computation of the sparse decomposition.Second,an improved adaptive sensing matrix is proposed,which is applied to the low frequency and high frequency wavelet transformation respectively.At last,OMP reconstruct algorithm is employed to reconstruct the wavelet transformation coefficients,and then the signal can be finally recovered through Wavelet synthesis.Simulation results demonstrate that,based on wavelet transform,the approach using the adaptive speech compressed sensing has a good performance in reconstruction.
出处
《南京邮电大学学报（自然科学版）》
北大核心
2012年第2期64-68,共5页
Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition
基金
国家重点基础研究发展计划(973计划)(2011CB302903)
国家自然科学基金(60971129)资助项目
