Adaptive signal processing with a self-development neural network

A self-developing neural network model for adaptive signal processing is proposed. The major improvement of this model over most current models is that the number of neurons can be dynamically adjusted by generating new neurons or annihilating existing nonactive neurons according to the training patterns. This model can be used to design adaptive signal processing systems with dynamically scalable processing capability. In particular, an adaptive source coder for data compression that is based on this neural network model is proposed