Radio Spectrum Sensing And Reconstruction Using Compressive Sensing

In the field of wireless communications, multiple signals normally coexist in oneband, such as WIFI, ZigBee and Bluetooth. Those wireless devices all operate in thelicense-free ISM2.4GHz band. Sharing band has increased the utilization of spectrumresources, and brought the risk of signal conflict at the same time. Different deviceswhich have similar center frequency working at the same time would make seriousinterferences to each other. Therefore, how to sense the utilization of spectrum in aregion has been an important issue.Utilization of spectrum is able to be characterized using received signal strengthindicator (RSSI). How to investigate the wireless spectrum usage in a specific spaceefficiently and conveniently is the focus of the study of this paper. The common wayof spectrum sensing are basically based on fixed point sensing, which is sensing everypoints’ data in the region and building regional the spectrum information map. Thereare other more accurate ways which need to get the specific modulation method andencoding format of a signal. Although those sensing ways have a very high accuracyrate, the time and effort costs of them are huge.This paper proposed a new spectrum sensing algorithm based on compressivesensing and matrix completion theory. This algorithm can restore the whole regionspectrum information by only collecting a few sparse spectrum data of this region.Thereby, it can effectively reduce the workload of data collection. Meanwhile, a newsoftware system has been designed to put the theory into practice by implementing theproposed algorithm. It also has integrated the functions of sensor implementation,data collection and data transformation.Finally, this paper proposed indoor and outdoor experiments in ISM band toverify the algorithm theory. Experimental results show that:1) In the case of occupiedindoor spectrum, spectrum information can be recovered less than standard error20by10%samples;2) In the case of idle indoor spectrum, spectrum information can berecovered less than standard error9by10%samples;3) In the case of idle outdoorspectrum, spectrum information can be recovered less than standard error2by10%samples.