Further Research On Blind Detection Algorithm Based On Chaotic Hopfield Neural Network

Compared with literature blind detection algorithms based on statistics, the Hopfield Neural Network(HNN) can be used to blindly detect BPSK signals with shorter received signals and it is appropriate for channels with common zeros, so it can satisfy the requirements that people proposed for the reliability of the wireless digital communication system. However, HNN is easily trapped in local minimum and even sometimes fail to get the globally optimal or near-optimal solutions due to the gradient descent dynamics. For this storage, this paper is supported by the National Natural Science Foundation of China(Grant No. 61302155), and introduces chaos technology in HNN and the main research work is as follows:(1) Due to the initial sending sequence generated by chaos has better ergodicity, this paper introduce chaotic initialization and chaotic disturbance to Hopfield Neural Network(HNN). The algorithm took the chaotic sequence generated by chaotic mapping as initial sending sequence, then introduced the chaos disturbance when the global optimal value trapped in a local minimum during a small range to reduce the error rate. Simulation results show that: the problem of quadratic optimization with constraints can be successfully solved with the HNN based on chaotic mapping,and the improved algorithm improves the algorithm performance.(2)For the phenomenon of premature occurred in the course of evolution of HNN as well as the slow convergence speed of Transient Chaotic Neural Network(TCNN), this paper proposes a novel network structure Improved Compound Sine Chaotic Neural Network(ICSCNN) based on TCNN, constructs a new energy function and proves the stability of ICSCHNN in asynchronous update mode and synchronous update mode respectively. Simulation results show that: compared with the second order statistics(SOS) algorithm、HNN blind detection algorithm、TCNN blind detection algorithm, the novel algorithm not only reduces the error rate markedly and convergence time, but also requires shorter data length, so that improves the performance of blind detection.(3)This paper introduces Wavelet Chaotic Neural Network(WCNN) into blind detection environment taking advantages of the higher nonlinear characteristic of wavelet that reference points out firstly, then adds a activation function for the each nerve cell to constitute Double Sigmoid Wavelet Chaotic Neural Network(DSWCNN), and constructs the novel network model according to characteristic of DSWCNN. Meantime, the stability of the network is also proved in asynchronous mode and synchronous update mode separately. Simulation results show that: the convergence speed of DSWCNN is faster than TCNN obviously,so DSWCNN improves the disadvantages of slow convergence speed of TCHNN well; meanwhile it has certain robustness for channels.