An Application of Improved RBF Neural Network in Modulation Recognition

In this paper, a novel modulation recognition method is proposed, which is based on an improved radial basis function (RBF) neural network. The parameters of radial basis function are obtained by fuzzy C-means (FCM) clustering algorithm, while weights of the network are trained with gradient descent approach. Optimal stopping rule is used to avoid overfitting and improve training speed as well as generalization ability. Application of this method to modulation recognition of practical signals shows satisfactory performance.