Greedy algorithms for sparse signal recovery based on temporally correlated experimental data in WSNs

Abstract

In wireless sensor networks (WSNs), thousands of sensor nodes are deployed to measure various environmental parameters such as temperature, light intensity, humidity and air pressure. All living beings can sense the variation in these parameters; therefore, these parameters are termed as natural signals. The natural signals are highly correlated in time and space; therefore, it can be compressed significantly to achieve the low sampling rates. The correlation property of natural signals is exploited here to compress/decompress these signals for reducing the transmission cost of the networks. The real-time temperature signal is measured using national instruments (NI) WSN platform, which is used for analysis purpose. The signal at first is transformed into sparse signal and then compressed. The compressed signal is transmitted to the receiver, where it is decoded into original sparse signal using algorithms based on greedy iterative approaches, i.e. orthogonal matching pursuit (OMP), stagewise orthogonal matching pursuit (StOMP) and generalized OMP (gOMP). The most popular greedy algorithm, OMP, is compared with StOMP and gOMP. The performance is analysed quantitatively in terms of peak signal-to-noise ratio, root- mean-squared error and execution speed of these greedy algorithms. It is demonstrated through simulation, the computational speed of StOMP and gOMP is much better than OMP, and also, the sparse signal is recovered with accuracy approximately equal to OMP.