Instantaneous frequency extraction of overlapped multicomponent signals with an adaptive constraints approach

Abstract Accurate extraction of instantaneous frequency (IF) from overlapped non-stationary signals is crucial in fields such as radar, communications, and biomedical engineering. Traditional approaches, such as ridge path regrouping (RPRG), rely on post-processing steps like truncation and interpolation, which may introduce artifacts and reduce robustness in complex signal environments. To address these limitations, an adaptive ridge path regrouping (ARPRG) algorithm is proposed. First, a variation-constrained time-frequency (TF) regulation filter is designed to suppress interference from other components. Then, an entropy-guided parameter adaptation mechanism is introduced to dynamically update parameters. These parameters are optimized using a genetic algorithm guided by local energy entropy, thereby eliminating the need for manual parameter tuning. In addition, a refined penalty term based on linear least squares fitting is employed to enhance estimation accuracy within overlapping TF regions. Simulation results and experimental validation on micro-Doppler radar signals confirm that ARPRG significantly improves IF estimation accuracy and robustness when compared to conventional techniques. This algorithm offers a robust solution for complex signal environments, with broad applications in advanced signal processing.