Blind Source Separation of Electromagnetic Signals Based on Swish-Tasnet

Digital devices may leak electromagnetic signals containing important information during operation, posing a risk of information leakage. In order to evaluate the safety level of the equipment, it is necessary to detect electromagnetic leakage signals and separate important information. Traditional methods such as ICA and IVA have limited performance when the signal input is smaller than the output. In recent years, signal separation technology based on deep learning has developed rapidly, demonstrating effective separation performance and potential in signal separation. For the separation of electromagnetic signals, we propose a blind source separation method for electromagnetic signals based on Swish-Tasnet. This method improves the time convolutional network structure and introduces the Swish activation function to optimize the internal functions and stacking layers of the model, thereby achieving efficient blind source separation of electromagnetic signals. The experiment evaluated the separation performance through SISDR, SDR, and SAR, and the results showed that this method performs excellently in reducing background noise interference and accurately separating signal sources from mixed signals. Swish-Tasnet not only provides a practical and effective technical means for electromagnetic signal leakage separation, but also brings substantial progress to the field of equipment information security assessment research. This method has broad prospects in practical applications and is worth further research and promotion.