MPSK Signal Blind Detection Based On Improved Amplitude And Phase Discrete Multilevel Complex Value Neural Network

Hopfield neural network means more than significant to the development of artificial neural network and it is especially important in solving optimization problems.The Hopfield neural network blind detection algorithm for multi-band signals has the advantages of having statistics independence and strong anti-jamming performance,which conforms to the requirements of modern communication systems.When dealing with multi-band signals,in the process of the Hopfield neural network find the optimal solution,it is easy to fall into the local optimum or miss the optimal solution after multiple iterations.In response to those problems,using the MPSK signal as the sending sequence,this paper made the following innovations:(1)The second chapter of this paper introduces the constant disturbance,self-disturbance and annealing disturbance on the basis of the model structure of Complex Hopfield Neural Network with Amplitude-Phase-type Hard-Multistate-activation-function.Given three kinds of the disturbed Complex Hopfield Neural Network with Amplitude-Phase-type Hard-Multistate-activation-function algorithm.At the same time,the network structure and dynamic equations of the new algorithm are given and proved the stability of the new network.The simulation results show that the proposed algorithm has stronger anti-interference performance and need less data under the common channels than the CHNN_APHM blind detection algorithm.The DCHNN_APHM algorithm with the self-disturbance performs best.(2)Aiming at solving the shortcomings of CHNN_APHM algorithm such as easy to fall into local optimum and multiple starting points,the third chapter introduces the transient chaos neural network(TCNN)into the MPSK signal blind detection and proposes the Complex Transiently Chaotic Neural Network with Amplitude-Phase-type Hard-Multistate-activation-function(CTCNN_APHM)blind detection algorithm.At the same time,a new network structure and corresponding dynamic equations are given.And proved the stability of the new neural network under synchronous model and asynchronous model mathematically.Experimental results show that the new algorithm has stronger anti-jamming performance than the common CHNN_APHM algorithm,the number of starting points needed for convergence is greatly reduced,and the data length required by the algorithm has decreased by 50%.However,due to the characteristics of TCNN's search capability,the time required for the convergence of CTCNN_APHM algorithm becomes longer,and the solution to the problem will be given in the follow-up.(3)The fourth chapter mainly solves the problem of slow convergence of CTCNN_APHM algorithm.This paper improves the convergence speed of the algorithm by improving the annealing strategy and presents two improved CTCNN_APHM algorithms based on piecewise annealing and logarithmic annealing.The experimental results show that the convergence speed of ICTCNN_APHM algorithm which has changed annealing strategy is greatly improved compared with that before the improvement,meanwhile,it also ensures the advantage of strong anti-jamming performance of the proposed algorithm in the common channel.Both annealing strategies performed well and solved the problems left in Chapter Three.