Abstract: Aiming at the two shortcomings of easy to fall into local minimum and inappropriate for small sample of neural network classifier, a new method of modulation recognition for digital signals is proposed using zero-center instantaneous features extraction to extract characteristics and based on support vector machine (SVM). Compared with traditional algorithms based on neural networks, this algorithm has better generalization ability. Computer simulation results indicate that the method has the high recognition rate with less samples.

Key words: modulation recognition, support vector machine, feature extraction