Intelligent Demodulation Method for Communication Signals Based on Multi-Layer Deep Belief Network

摘要

Aiming at the problem of signal demodulation under noise interference channel, a signal recognition method using deep learning is proposed. The signal demodulation is completed by identifying the signal. The deep confidence network uses a restricted Boltzmann machine as the basic unit to design a multi-layer deep confidence network for communication signal identification. The communication signal is first transformed into a specific characterization sequence, and a complete training set is constructed to perform layer-by-layer unsupervised learning and global supervised fine-tuning feedback learning for the deep confidence network, and realize communication in the weight parameter optimization process of the deep confidence network. Feature extraction and recognition of signals. Simulation experiments show that compared with the traditional modulation signal demodulation method, the detection performance of the signal demodulation method using deep learning is about 0.4 dB boost.