A NONLINEAR SPECTRUM ESTIMATION SYSTEM USING RBF NETWORK MODIFIED FORSIGNAL-PROCESSING

This paper proposes a nonlinear signal processing by using a three lay ered network which is trained with self-organized clustering and super vised learning. The network consists of three layers, i.e., a self-org anized layer, an evaluation layer and an output layer. Since the evalu ation layer is designed as a simple perceptron network and the output layer is designed as a fixed weight linear node, the training complexi ty is the same as a conventional one consisting of self-organized clus tering and a simple perceptron network. In other words, quite high spe ed training can be realized. Generally speaking, since the data range is arbitrary large in signal processing, the network should cover this range and output a value as accurately as possible. However, it may b e hard for only a node in the network to output these data. Instead of this mechanism, if this dynamic range is covered by using several nod es, the complexity of each node is reduced and the associated range is also limited. This results on the higher performance of the network t han conventional RBFs. This paper introduces a new non-linear spectrum estimation which consists of LPC analysis and RBF network. it is show n that accuracy spectrum envelopes can be obtained since a new RBF net work can estimate some nonlinearities in a speech production.