The extraction of latent fine feature of communication transmitter

In order to fully exploit the fine features of transmitter in the high dimensional feature space, we propose a method for extracting the fine features of communication transmitter based on global latent low rank representation. Firstly, we extract the instantaneous frequency of the signal of communication transmitter, and then transform the signal into a high dimensional feature space by the Fourier transformation. Secondly, the global low-rank structure between samples and global latent low-rank relationship between the dimensions are mined. Therefore, we can put the whole training sample set in the latent low rank representation model and get the latent part matrix of training sample set based on the low rank relationship between feature dimensions. Every column vector of latent part matrix is the latent fine feature vector of each communication transmitter signal. On the actually collected data set of the FM radios with the same model and manufacturer, the latent fine features extracted by our method can significantly improve recognition performance of different transmitter radios which reflects the effectiveness and robustness of our method.