Deep Interference Recognition for Spread Spectrum Communications using Time-Frequency Transformer

Abstract

Spread spectrum techniques have been widely adopted in both military and commercial communications because of their anti-interference and anti-interception properties. How-ever, when the interference power is high and the spread spectrum gain is not sufficient to achieve satisfactory system performance, additional anti-interference techniques need to be employed, and the design of the interference sensing and recognition method is a basic prerequisite of effective interference suppression. In this work, considering the superiority of transformer networks in the field of deep learning, we propose a novel interference recognition method with short time Fourier transform (STFT) analysis and transformer. For making full use of the hidden information in both time and frequency domain, the proposed method intro-duces the STFT analysis to extract the high-dimensional feature. Furthermore, the idea of transformer is adopted in our method to introduce the attention mechanism focusing on the time-frequency correlation of the received signals and improve the recognition accuracy. Simulation results demonstrate the superiority of the proposed method versus other baseline competitors.