The Study Of Blind Detection Algorithm Of Underwater Discrete Multi-level Hopfield Neural Network

The technology of underwater communication plays an important role in the fields of national maritime military strength construction and the exploration of marine resources.As competition between countries becomes more and more fierce,underwater communication technology has become a hot research field.Realizing blind detection of signals in underwater digital communication systems has great significance in underwater communication technology.The blind detection algorithm based on Hopfield neural network(HNN)is widely used in digital communication system,which can detect the received signals directly without relying on statistics,and improves the optimization performance of blind detection algorithm by using non-linearity and self-adaptability of neural network.It has been proved in literature that the blind detection algorithm of Complex-valued Hopfield Neural Network Real-Imaginary-type Hard-Multistate-activation-function(CHNN_RIHM)can realize blind detection of discrete multilevel signals in wireless communication system.Therefore,this paper attempts to apply CHNN_RIHM algorithm in underwater digital communication system,and proposes Underwater CHNN_RIHM(UCHNN_RIHM).However,blind detection algorithm based on Hopfield neural network is easy to fall into local minimum by gradient descent method when searching the global optimal solution,and also needs large data length,as well as high signal-to-noise ratio.These problems limit the application of UCHNN_RIHM in underwater digital communication systems.In response to above problems,the following innovations for UCHNN_RIHM algorithm has been done:(1)Aiming at the shortcoming of UCHNN_RIHM algorithm easily falling into local minimum,the second chapter introduces Transient Chaotic Neural Network(TCNN)to UCHNN_RIHM algorithm,and constructs a blind detection algorithm model based on underwater chaotic neural network.TCNN utilizes chaotic attractor to realize the global search in order to find the global optimal solution,while avoiding the algorithm falling into local minimum.At the same time,this chapter introduces hysteresis noise in new model to improve the anti-noise ability of the blind detection algorithm,and constructs the blind detection algorithm of Underwater Complex-valued Transiently Chaotic Neural Network Real-Imaginary-type Hard-Multistate-activation-function(UTCNN_RIHM).The dynamic equation of UTCNN_RIHM algorithm is given in this chapter.The new energy function is given and the stability of algorithm is proved.Finally,the experimental results show that the UTCNN_RIHM algorithm not only reduces the length of data but also improves the anti-noise ability while avoiding falling into the local minimum.(2)For TCNN,the linear annealing function is used as the self-feedback,which takes a very slow cooling time to obtain the global optimal solution,leading to the slow convergence of UTCNN_RIHM algorithm.The third chapter introduces the segmented annealing function as self-feedback instead of the linear annealing function to improve the convergence speed of UTCNN_RIHM algorithm.Then,a blind detection algorithm called Underwater Complex-valued Transiently Chaotic Neural Network Based on Stepped Annealing Real-Imaginary-type Hard-Multistate-activation-function(UTCNNS_RIHM)is proposed in this chapter.The experimental simulation results show UTCNNS_RIHM algorithm with segmented annealing function has a faster convergence speed and stronger anti-noise ability than UTCNN_RIHM algorithm.(3)The active function in UTCNNS_RIHM algorithm processes the real and imaginary part of signal,then integrates the processed results into a complex signal as a neuron state,resulting in cumbersome algorithm,high complexity,and slow convergence speed.In the fourth chapter,by referring to clustering idea of K-means algorithm,the multi-clustering activation function is proposed to processe the signal as a whole,and the state of neuron is obtained after the new activation function.Then,the Blind Detection Algorithm Based on Multi-Clustering Underwater Transiently Chaotic Neural Network(MCU_TCNN)is proposed.The experimental results show that the MCU_TCNN algorithm has certain advantages over UTCNNS_RIHM algorithm in terms of anti-noise ability,and improves the convergence speed of the blind detection algorithm,making the blind detection algorithm more widely applicable at low Signal-to-Noise Ratio(SNR)and small data length.