Research On Blind Detection Algorithm Based On Improved Hopfield Neural Network

The blind detection algorithms of Hopfield Neural Network(HNN) is independent of the statistics information of the received signals. They are applicable in the cases of the channel with common zeros and the short data block. Compared with the second order statistics(SOS) algorithm and high order statistics(HOS) algorithm, the HNN algorithms can meet the transmission requirements of current high-speed communication system. But in the existing literatures,HNN seems easily trapped in local minimum and even fails to get the globally optimal, which affects the anti-interference ability and convergence speed of the algorithms. In order to improve the performance of blind detection system, the main contributions of this paper are as follows:(1) The Variable Step Hopfield Neural Network structure(VSHNN) is based on the proposed HNN model which applies for blind detection in the communication system. The new activation functions are constructed and the stabilities for VSHNN with asynchronous and synchronous operating mode are also analyzed respectively. Simulation results show that the blind detection algorithm of VSHNN is suitable for the channel with common zeroes, and the required data size is very short. Compared with the HNN?TCNN?DS-TCNN algorithms, the VSHNN algorithm has better anti-interference ability via some channels. In addition, the VSHNN algorithm is superior to the HNN and TCNN algorithm in terms of the convergence performance, but inferior to the DSTCNN algorithm.(2) Compared with other Hopfield neural network blind detection algorithms, the convergence speed of the DS-TCNN algorithm is obvious owing to its own characteristics, while its antiinterference ability is defective. In order to improve the anti-interference ability of DS-TCNN algorithm, a new DS-DCNN algorithm based on DS-TCNN is proposed in the third chapter. In this new algorithm, we use three different disturbance terms and construct a new energy function. Furthermore, the DS-TCNN algorithm proves to be stable in both synchronous and asynchronous update modes. Finally the simulation results show that the novel algorithm not only significantly reduces the error rate and convergence time, but also requires the less amount of data size compared with the DS-TCNN algorithm?HNN algorithm and SOS algorithm.(3) To further enhance the DS-DCNN convergence speed, the DS-NSCNN is constructed in the fourth chapter. We use three different nonlinear self-feedback terms for the double sigmoid chaotic neural network structure, then the DS-NSCNN blind detection algorithm is put forward. Firstly, we constructs the DS-NSCNN model and design the network energy function. Sencondly, the DSNSCNN algorithm proves to be stable in both the synchronous and asynchronous update modes. Finally the simulation results show that DS-NSCNN algorithm inherits the advantages of DS-TCNN. This proposed algorithm requires only a small amount of data to achieve the blind detection. In addition, the DS-NSCNN algorithm has better anti-interference ability and energy function convergence rate than the VSHNN algorithm, DS-DCNN algorithm and some literature algorithms.