Blind multi-valued signals detection using discrete Hopfield network
Yun Zhang, Zhiyong Zhang · 2010 Sixth International Conference on Natural Computation · 2010
The conventional neural networks which are limited to two-state neurons are not able to solve the problem of blind multi-valued signal detection. A new algorithm based on discrete Hopfield neural network(DHNN) is proposed to detect multivalued signals blindly. A discrete 4-level signum-type activation function is constructed for 4PAM signals. For the blind signal detection, the optimization performance function is constructed and it does not rely on the second or higher order statistics of the received signals. Based on the new weight matrixes and the energy function of multi-value DHNN, the stability for multi-value DHNN is also proved in the paper. Simulation results show that the algorithm reach the real equilibrium points in a few iterations and show high speed to blindly detect multi-valued signals in stochastic channels.