A Power-Sensitive Feature Index Model-Based Method for EMC Prediction of GPS Receivers

Global navigation satellite system receivers are highly vulnerable to complex electromagnetic interference, while existing studies face difficulties in quantifying interference parameters and establishing a unified framework across different interference types. To address this challenge, this article proposes a receiver anti-interference performance evaluation method based on a power-sensitive characteristic exponential model. The approach first eliminates spectral amplitude deviations through L2-norm normalization, and then quantifies the similarity between measured and reference spectra using cosine similarity. A power penalty term is introduced to construct a weighted similarity measure, which is further combined with dynamically updated frequency weights to fuse multifrequency information and yield interference-sensitive feature scores. A joint loss function is designed to iteratively optimize the frequency weights and a unified decision threshold. Finally, discrete scores are fitted with an exponential model to form continuous power-sensitive characteristic exponential models. Experimental results demonstrate that the optimized unified threshold achieves reliable failure detection, with fitted threshold powers deviating within$\pm 3$dB of measurements. Additional validation with signals of varying pulse duty cycles confirms the generality of the method.