基于同伦算法的时变流信号动态压缩感知研究

Study on Dynamic Compressed Sensing Based on the Homotopy Algorithm to Process Streaming Signals

为解决传统压缩感知在处理时变流信号时易引发的效率低、块效应、延迟高等问题,提出了基于L1同伦恢复算法的时变同伦算法。该算法通过对重叠的信号时间采集窗口的滑动控制,压缩采样当下窗口中的待测数据,并用L1同伦恢复算法实时恢复出原信号。仿真表明,以医学领域的心电信号为例,该算法能够快速实时地压缩采样并精确重构时变心电信号,其重构信号的相对误差控制在10-2范围内,充分证明了算法的实践可行性,从而能有效地解决存储庞大动态心电图数据的难题。

Currently most of the compressed sensing to date has focused on static finite-dimension signals.The real-time homotopy algorithm based on the L1-homotopy recovery algorithm was proposed.The method iteratively processed measurements over a sliding overlapping window and chosed the L1-homotopy recovery algorithm that could avoid solving a new L1 program from scratch at the next iteration to solve a weighted L1-norm problem for estimating sparse coefficients.Experiment results on very long real-time electrocardiosignals in the field of medicine demonstrated the good performance of the algorithm in real-time system,with the relative error of the reconstructed signal controlled within 10-2,which effectively lessened great storage burden the ambulatory electrocardiograph(ECG) had brought about.