Blind source separation and tracking of multiple frequency hopping signals for cognitive radio communications

In this paper we consider a blind source separation problem in which the number of sources is unknown and all or some of the sources may be networks with frequency hopping spread spectrum capability. We should first find the number of sources and then fairly track each of the detected networks. The extracted information of detected sources are used to update an internal model in the cognitive radio that helps it to predict and avoid frequencies in which active FHSS radios in the environment are simultaneously transmitting. In the proposed method, there is no need to have several antennas for DOA estimation and this will significantly reduce implementation cost and complexity of the algorithm. Due to real time nature of the problem, the simple structure of this method in comparison with other existing algorithms will make it preferable. We propose a method based on a probabilistic classifier to separate detected networks in sequential frames. The efficiency of the proposed method is demonstrated by simulation results.