Sub-band Selection for Transform Domain Communication System Based on DRQN

Spectrum sensing is a prerequisite technology for transform domain communication system (TDCS). Due to technological limitations, the working sub-band of TDCS is randomly selected. In order to improve the anti-interference ability and low interceptability, we designed a technique for dynamically selecting sub-band. In this paper, we model the problem of dynamically selecting sub-band of TDCS as a partially observable Markov decision process (POMDP) and propose an efficient sub-band selection strategy based on deep recursive Q network (DRQN). It is shown that the proposed DRQN-based working sub-band selection strategy has low computational complexity and higher accuracy than random sub-band selection strategy.