Blind Detection of Non-cooperative Communication Signals Based on Feature Dimensionality Reduction

Due to the complexities of channel environments and the diversity of signal types, blind detection of communication signals in a non-cooperative communication environment remains a challenging problem. To alleviate these issues, we proposes a blind detection method for communication signals based on feature dimensionality reduction. Initially, multimodulus algorithm (MMA) is employed to eliminate the effects of the channel and a short time Fourier transform (STFT) is applied to generate a time-frequency representation. Furthermore, a combination of convolutional neural network (CNN) and linear discriminant analysis (LDA) is utilized to extract common features that distinguish multiple types of signals from noise. This approach significantly enhances the algorithm's ability to extract features and improves detection rates. Experimental results demonstrate that, in a Rayleigh channel environment, with a false alarm rate of 10
^%
and a signal-to-noise ratio of -10dB, the proposed algorithm achieves a detection rate of 94 %, which represents a notable improvement of approximately 10% compared to the STFT-CNN algorithm.