Power Spectrum Detection Using Clustering

Spectrum detection is the basic tool to permit cognitive radio to utilize an empty channel and opportunistically transmit. Considering the sparse utilization of the frequency spectrum, in this paper we propose the use of k-means clustering algorithm to create an sparse representation of the Power Spectrum Density (PSD) of a received signal, and a method to extract the spectral information from it. Preliminary results show the possibility of to identify the occupied channels using this sparse representation followed by some simple processing. The proposed method have low complexity, and under proper conditions it can achieve approximately 99% of correct channel detection on average.