A 1-Dimension Structure Adaptive Self-Organizing Neural Network for QAM Signal Classification

The 1-dimension structure adaptive self-organizing neural network (1-DSASONN) has been presented as an extended 1-dimension version of the self-organizing map, which has better performance in a modulation classification method proposed in this paper for QAM signals. 1-DSASONN can start with arbitrary number of neuron, grow or prune many neurons and adoptively adjust network structure as well as weights. This feature improves the efficiency of classification algorithm which utilizes the number of sets of equal amplitude as classification feature. Simulation results show that the modulation classification method is robust in the presence of phase estimation error.