Abstract: With the analysis of the pattern recognition based on sparse representation,a new feature extraction method using K-SVD and sparse representation was proposed to improve the accuracy of the digital modulation recognition under the low signal-to-noise ratio．Firstly,the principle component analysis was put forward to reduce the dimensionality of the samples．Secondly,the sparse dictionary was constructed by the algorithm of K-SVD．Finally,the sparse representation of the sample was calculated by ¹-minimization,and the feature was extracted according to the distribution of the sparse coefficient values．The identification problem was solved by using SVM classification machine．The simulation results indicate that the performance of this feature values extracted by this new algorithm is feasible in engineering application.

Key words: Modulation recognition,Sparse dictionary,Sparse representation,Support vector machine