An adaptive compressed sensing method based on selective measure

Abstract

An adaptive compressed sensing architecture based on selective measure is proposed in this paper, in order to reduce the effects of non-sparse noise component on the performance of existing compressed sensing reconstruction algorithm. Firstly, in this paper we analyze and deduces the statistics characteristic of the measured noise and its influence on the reconstruction performance; then we propose a compressive-domain projection filter combined with iterative noise detector method to obtain the location information of noise subspace based on minimal output energy criteria; thirdly, we measure matrix adaptively with the location information, and focus on the signal subspace directly without sensing the noise component in analog part. Simulation results show that compared with the existing compressed sensing procedures, our method can obviously improve the performance of reconstruction of signals with noise, and can be used to perform the front-end spectrum analysis of absorbing materials and better detect the active channels in cognitive radio.