Wave-shape function model order estimation by trigonometric regression

The adaptive non-harmonic (ANH) model is a powerful tool to compactly represent oscillating signals with time-varying amplitude and phase, non-sinusoidal morphology. Given good estimators of instantaneous phase we can construct an model, where the morphology oscillation described by wave-shape function (WSF), 2{\pi}-periodic more general periodic function. In this paper, address problem estimating number harmonic components WSF, that remains underresearched, adapting trigonometric regression selection criteria into context. We study application these criteria, originally developed in context stationary signals, case amplitudes phases. then incorporate order estimation ANH reconstruction procedure analyze its performance for noisy AM-FM signals. Experimental results on synthethic indicate enable waveform non-stationary oscillatory patterns, even presence considerable amount noise. also apply our task denoising simulated pulse wave determine proposed technique performs competitively other schemes. conclude work showing algorithm takes account interpatient variability electrocardiogram (ECG) respiratory analyzing recordings from Fantasia Database.