Blind Detection Of MPSK Signal With Complex Hopfield Neural Network

Blind detection is one of the latest developments because it can detect signals without training sequence. This thesis introduces the foundational principles and development of blind detection techniques based on neural network and analyzes blind detection algorithms of Hopfield neural networks.This thesis proposes a complex Hopfield neural network to blindly recover MPSK signals. This algorithm has following distinguishing features: Based on the relationship between received signals and projection operator of transmitted signals, the thesis will construct new optimization performance functions to blindly detect signals. It also will construct complex-value activation function based on a polar coordinate. The weight matrix based on projection operator of the received signals will be configured. The optimal points of a constrained quadratic programming performance function for blindly detection of signals are mapped to global minimal points of the new energy function.Simulation results show that the complex Hopfield neural network can quickly converge to the optimal point with a very short received sequence, and blindly detect signals with a low bit error rate.