压缩感知框架下非均匀信息采集及重构

Non-uniform information acquisition and reconstruction within compressed sensing framework

为提高现有直接信息均匀压缩采样(Analog to information conversion,AIC)的观测效率和性能,在能量准则下,提出非均匀信息采集(Non-uniform information acquisition,NUIA),充分利用信息的重要性先验,即对信号随机调制后,依据能量进行变速率的采集,能量越大的信号段采样速率越高,反之亦然。结合支撑域扩充、剪枝的思路提出变速匹配追踪(Variable rate matching pursuit,VRMP)算法,通过引入非均匀观测的先验支撑集,并在追踪过程中将其与迭代估计出的支撑集相并,提高了重构精度。理论分析和实验结果表明了NUIA-VRMP的有效性。特别地,相比于常规AIC的子空间追踪重构,NUIA-VRMP的组合能在低采样速率条件(如20%的Nyquist速率)下获得50dB的重构增益。

The existing Analog information measurement, and Under the energy criterion, a After random modulation, the higher Variabl the sampling rate is an to Information Conversation （AIC） is based on uniform-low-rate the importance prior information contained in the signal is underused. Non-Uniform Information Acquisition （NUIA） method is proposed. signal is sampled at non-uniform-rate that the bigger the energy is the d vice versa. Combing the idea of support merger with pruning, a e Rate Matching Pursuit （VRMP） algorithm is pr united with the set of Compared with the Subs NUIA-VRMP can obtain signal appropriation support, pace Pursuit （SP） reconstruc 50dB reconstruction gain at ultra oposed. The prior support set, which is able to promote the recovery accuracy. of conventional AIC, the combination of low-rate, （e. g. 20% Nyquist Rate）.