Abstract:

In order to solve the dilemma of the tradeoff between spectrum sensing performance and spectrum sensing efficiency in cognitive radio network，a nove1 ED/FD cooperation spectrum sensing algorithm based on featured belief points was proposed. Firstly，this algorithm desployed the single cognitive user energy detection and feature detection co11aborative detection mode instead of multiple cognitive user cooperative detection，reducing communication overhead. Secondly，it modeled cognitive radio network under dynamic uncertainty using partia11y observable Markov decision processes（POMDP），and transformed the optimization of the tradeoff between sensing performance and sensing efficiency into yielding the optima1 value function of POMDP. Fina11y，a nove1 approach using characteristics of belief to contro1 the scale of belief states was presented，which exploited the maximum online reward value iteration algorithm to approximate the optima1 value. The numerica1 results show that the proposed algorithm is able to meet the requirement of high tracking performance with constraint of 1ow probability of interfering primary users.

Key words: cognitive radio network, spectrum sensing, energy detection, feature detection, partia11y observable Markov decision process