Demodulator based on deep belief networks in communication system

Abstract

Deep belief network (DBN) has been successfully applied in variety areas such as image recognition and natural language processing. In this paper, we investigate the signal demodulation problems in different types of communication channels. Then, a novel deep belief networks (DBN)-based demodulator is proposed. Since the DBN-based method is just like a black box that can automatically learn how to demodulate the received signals, few manual designs is required in the receiver. Moreover, we also propose a novel mapping method for the communication signals to match the input of DBN. Simulation results with different amounts of training samples and iterations show that the DBN-based demodulator is feasible and efficient.