A time-varying parameter estimation method for oscillometry respiratory impedance measurement based on time–frequency characteristics separation

Accurate estimation of time-varying respiratory impedance in multi-frequency oscillometry is essential for dynamic lung function assessment, but remains challenging because of signal superposition, frequency leakage, and the limited temporal resolution of conventional spectral methods. This study proposes a signal-processing and recursive identification framework for tracking respiratory impedance within the clinically relevant frequency range of 4–50 Hz. The framework first uses a zero-phase time–frequency separation filter to decompose composite oscillometry signals into frequency-specific components while minimizing phase distortion and spectral leakage. Based on the separated signals, recursive least-squares identification strategies are then developed to estimate frequency-dependent impedance trajectories using both a linear time-varying model and a generalized frequency-domain model for time-varying effects. The proposed method was evaluated using simulated data and human oscillometry measurements. Compared with existing filtering and sliding-window spectral methods, the proposed filter achieved more accurate signal separation with negligible frequency leakage, and the overall accuracy of time-varying impedance estimation improved by at least 30%. These findings demonstrate that the proposed framework provides an effective and robust solution for within-breath respiratory impedance tracking, with improved temporal resolution and estimation accuracy for multi-frequency oscillometry analysis.