Research on adaptive signal detection and feature extraction method of complex electronic information system

Abstract

Aiming at the challenge of signal detection and feature extraction in complex electromagnetic environment, this paper proposes an adaptive joint optimization method of signal detection and feature. Faced with the problems of traditional methods relying on prior knowledge, insufficient deep learning generalization ability, and poor adaptability to non-stationary noise, a cascaded framework was constructed that includes variable step size LMS (VSS-LMS) interference suppression, multi head self-attention (MHSA) feature extraction, and dynamic threshold decision-making. VSS-LMS achieves fast convergence and steady-state error balance by improving the tongue line function. MHSA uses multi-scale attention mechanism to focus on the characteristics of target signal adaptively. Adaptive threshold dynamically adjusts the threshold based on local noise power estimation. Experiments show that the detection probability of this method is 96% in the signal-to-noise ratio range of -15 dB to 5 dB, which is more than 15% higher than that of the traditional method. The accuracy of feature classification is over 94.3%, and the false alarm probability is stable below 0.01. The three modules cooperate to improve the detection sensitivity and feature discrimination in dynamic electromagnetic environment, and provide a new scheme for intelligent perception of complex electronic information systems.