Compressed Sample-based Spectrum Sensing Algorithms

Cognitive radio is proposed to solve low usage efficiency of authorized spectrum. It focuses on how to endow wireless devices with the capability of sensing'spectrum holes'and utilizing them without interference on authorized users. Cognitive radio is based on the following fact. On a certain time, only a minority of spectrum bands are occupied by authorized users. In this sense, spectrum signals sensed by cognitive users can be regarded as sparse signals.At the same time, as an emerging research filed, compressive sensing relies on sparsity of underlying signals. Through random measurement matrices, a high-dimensional signal can be mapped to a much smaller number of linear measurements with deterministic performance guarantees.In practical wireless environment, a cognitive network is always formed by many independent cognitive radio users. As an independent cognitive radio user, accurate sensing result can not be achieved because of link attenuation, multi-path effect, thermal noise, diversified random interference etc. So this paper aims at distributed cognitive network and utilizes compressive sensing to sense spectrum status. Through cooperation between cognitive radio users, accuracy of sensing result can be elevated and complexity decreased.To solve different problems, several different algorithms are proposed using compressive sensing in distributed cognitive network. To overcome link attenuation and noise effect, a centralized algorithm is proposed to make a single compressive sensing shared by all cognitive radio users. The sampling number of every user is proportional to its radio environment. This algorithm leads to such problems as synchronization and communication overload problems. Based on these problems, this paper proposes a distributed compressive sensing algorithm. Cooperation among users is achieved though mutual information evaluation. Finally, combining the above two algorithms, this paper proposes a distributed algorithm instead of centralized algorithm using factor graph to make a single compressive sensing shared by all users. Simulation shows every algorithm shows a better performance.