Research On Blind Detection Algorithm Based On Improved Hopfield Neural Network

Compared with the traditional blind detection algorithm,the blind detection algorithm based on Hopfield neural network(HNN)does not depend on the statistics and is not limited to channels containing common zero,which has the irreplaceable advantages.However,with the development of Internet of Things and large dataset,and the arrival of 4G and 5G era,modern communication systems have higher requirements for the reliability and validity of signal transmission.This paper aimed at these two factors,mainly to make the following innovative works:(1)In order to improve the anti-interference ability of Hopfield neural network,and improve the poor mapping ability,insufficient nonlinearity,low flexibility of the traditional activation function,and to solve the higher sensitivity of the neurons near the zero point,an improved adjustable sigmoid function Hopfield Neural Network(ASHNN)blind detection algorithm is proposed in the second chapter.With the advantages of adjusting the steepness,position and mapping range at the same time,the novel activation function can reduce the derivative value of the activation function,and reduce neuronal sensitivity to input values near zero,so improve the anti-interference ability of the system.Experimental results show that the flexibility of the improved activation function,the self-organization ability of learning ability of HNN,and the error performance and convergence speed of blind detection algorithm are greatly improved.(2)In order to improve the vulnerability of Hopfield neural network to local optima and the slow convergence of hysteresis chaotic neural network(HCNN),based on Transient Chaos Neural Network(TCNN),a double sigmoid hysteresis chaos neural network(DS-HCNN)is proposed in the third chapter.We firstly design a new hysteretic activation function and study the selection of its parameters in detail,then we constructed a new energy function and proved its' stability.The final experiment proves the effectiveness of DS-HCNN algorithm from multiple perspectives such as bit error rate(BER),data length,convergence time and distance norm.(3)In order to further improve the performance of the algorithm,on the basis of the third chapter,we propose a double Sigmoid Hysteresis Noise Chaos Neural Network(DS-HNCNN)blind detection algorithm in the fourth chapter: the introduction of random noise perturbation to make the network having the characteristics of random chaos simulated annealing;the strategy of improved random simulation annealing to reduce the time complexity.Simulation results show that the improved DS-HNCNN algorithm further optimizes the DS-HCNN algorithm.DS-HNCNN shows stronger error performance under the same initial conditions,and what's more,the data length of the transmitted sequence is shorter and can be very suitable for classical channel,the convergence speed has been further improved,which successfully implement signal blind detection.