Cumulant-Based RLS Algorithm with Variable Forgetting Factor to Estimate Time-Varying Interharmonics

In this paper, an improved recursive least square (RLS) algorithm was proposed to estimate time-varying AR parameters in the presence of noise. Interharmonics signal can be modeled as a nonstationary auto-regressive (AR) model, the spectral estimation of interharmonics signal can be given by the estimated time-varying AR parameters. AR parametric spectral estimation methods have better frequency resolution. However, the conventional RLS algorithm is sensitive to noise, and fixed forgetting factor (FFF) has poor adaptability in the nonstationary environment. A new mean-squared-error (MSE) objective function based on fourth-order cumulant was introduced in this paper, which can suppress the Gaussian noise. For estimating the time-varying spectra of nonstationary signals using variable forgetting factor (VFF). The results of simulation proved that in noisy environment, this proposed method can get the spectral estimation of time-varying interharmonics accurately.