Modulation Classification of Linear Digital Signals Based on Compressive Sensing Using High-Order Moments

Compressed sensing theory can be applied to reconstruct the signal with far fewer measurements than what is usually considered necessary. While for the classification of modulated signals, we only expect to acquire some characteristics rather than the original signal. However, to select the feature used for modulation classification with sparsity is the main challenge. In this paper, we propose a method to identify the linear modulation format of an unknown single carrier linear digital signal using compressive samples, without reconstructing the original signal. In our method, we construct a compositional feature of multiple high-order moments of the received data as the identification characteristic. From simulations we can see that the method is effective, even at a relatively low signal-to-noise ratio.