多尺度自适应直接信息采样与重构

现有的直接信息采样(Analog-to-Information Conversion,AIC)方法在压缩感知框架下,将采集和压缩过程融合,但并未充分考虑不同稀疏成分在信号重构当中的地位。针对该不足之处,提出一种多尺度自适应直接信息采样与重构算法。该算法充分考虑小波变换后的高频信号和低频信号在重构中的不同地位,实现速率自适应的直接信息采样。同时,给出变速率采样下的信号重构策略,以解决常规重构算法在速率自适应采样时失效的问题。仿真结果表明,多尺度自适应AIC系统可以获得比传统AIC系统更好的AFSNR性能。

Multiscale Adaptive Analog-to-Information Conversion and Reconstruction

In the analog-to-information conversion (AIC) system within the compressed sensing framework, the sparsity ofinput plays an important role on its precise reconstruction. But most signals in our daily lives are not absolutely sparse butapproximately sparse or compressible ones, so a method of multiscale adaptive analog-to-information conversion and re.construction was proposed, and the validity was proved in this paper. By introducing orthogonal wavelet matrix (OWM) inthe recovery end equivalently, multiscale adaptive AIC considered the different roles of coarse coefficients and detail onesin the reconstruction after wavelet transform and realized the rate adaptive sampling. Meanwhile, the adaptive orthogonalmatching pursuit (AOMP) algorithm was proposed to solve the problem that the conventional reconstruction algorithm failsin rate adaptive sampling. The simulation results show that the improved AIC system could get better AFSNR performancethan conventional AIC system.