Abstract: Aiming at the problem of feature parameter numerousness and feature redundancy in the recognition of digital modulation signals, a feature optimization method for digital modulation signals recognition was proposed in this paper.Firstly,the method optimized the feature parameters of twenty selected features by orthogonal experiment And then it recognized nine kinds of digital modulation signals using RI3F neural network. Finally, the method compared to PCA method and KPCA method. The simulation results show that the method is able to optimize feature parameters effeclively in Gaussian and multipath channel,and has much better optimization ability than PC八and KPCA methods.

Key words: Modulation recognition, Feature parameter selection, Orthogonal design experiment, PCA, KPCA