Feature classification for the satellite modulation based on sparse coding algorithm

In order to improve the characteristics for satellite data expression ability, it is proposed sparse coding with over complete bases, to induce the high dimensional feature vectors form down to up pattern, then to accurately express the original high-dimensional features with very few over complete basis vectors. From top-down semi supervised learning characteristics, to project high dimensional feature to low dimensional space, in order to verify the similarity of coding of the training data, then to express the characteristics of input satellite data. The encoder needs as much as possible reconstruction for the input multi dimensional, and the find the main ingredients for representing the original information about input data. Therefore, the over complete is not only by the coefficient of input data to determine characteristics, but also by the dimensional space. In order to verify the performance of the sparse coding algorithm, using the sparse coding algorithm is to identify the 6 kinds of commonly used digital modulation signal: 4ASK, 4FSK, 4PSK, MSK, 16QAM, π/4QPSK. The performance of correct modulation recognition rate is higher stability. The overall recognition of 6 kinds of modulation rate of SNR is higher, not less than 0dB 99%.