A Novel Blind Detection Algorithm Based on Multi-Clustering Complex Neural Network

Abstract

Aiming at solving the problems that the CHNN-RIHM (Complex-valued Hopfield Neural Network Real-Imaginary-type Hard-Multistate-activation-function) is easy to get into a local optimization and requires a large amount of data, we propose to use Signal Blind Detection Algorithm Based on MC-CHNN (Multi-Clustering Complex-valued Hopfield Neural Network), construct a new energy function and prove the stability of MC-CHNN. The optimization scheme of MC-CHNN is as follows: we use a piecewise annealing function to improve the convergence speed of blind detection. To further reduce the complexity of the MC-CHNN algorithm and improve the sensitivity of MC-CHNN, we propose to apply a new activation function called the multi-clustering function to MC-CHNN and replace the activation function of CHNN-RIHM with multi-clustering function when dealing with discrete multilevel signals. The simulation results show that the MC-CHNN has a faster convergence speed, stronger anti-noise capability, and can be better applied to low SNR than its competitors.