摘要
在研究语音信号压缩感知时,观测矩阵是数据采样和信号重建的关键。一个好的观测矩阵不但可以使原始语音信号投影到一个低维空间上,而且可以保证该信号不丢失有用的信息,最终通过重构算法实现精确重构。文章比较几种常见的观测矩阵在语音信号压缩重构过程的性能,并对在不同条件下观测矩阵的选取提出建议。实验结果表明:压缩比和帧长是观测矩阵选取的关键因素;在不同的压缩比和帧长下,需要选取不同的观测矩阵,以达到最好的语音重构效果。
In the study on compressed sensing of speech signal, the observation matrix is very important for data sampling and sig- nal reconstruction. A useful observation matrix can make the original speech signal to be projected onto a lower dimensional space with all the useful information of the signal. And the signal can be reconstructed with proper algorithms. This paper introduced the theory of speech signal compression and compared several common performance indexes of observation matrix. Finally, it was suggested how to select the observation matrix under different experimental conditions. The experimental results show that compression ratio and frame length are key factors in selecting observation matrix. Selecting observation matrix according to different compression ration and frame length will be helpful to achieve better speech reconstruction effect.
作者
强策
夏凌
李光瑞
QIANG Ce XIA Ling LI Guangrui(School of Electrical Engineering and Electronic Information ,Xihua University, Chengdu 610039 China)
出处
《西华大学学报（自然科学版）》
CAS
2017年第2期24-27,共4页
Journal of Xihua University:Natural Science Edition
基金
教育部春晖计划项目(Z2015113)
四川省科技支撑计划项目(2015JY0138)
关键词
语音信号
压缩感知
观测矩阵
压缩比
帧长
speech signal
compressed sensing
observation matrix
compressionratio
frame length
