Abstract: Window-based estimates for stochastic harmonic regression models are useful for cases where harmonic parameters appear to be time-varying. Least squares estimates for harmonic models with one fundamental have been studied and asymptotic variance expressions have been developed. This paper extends these results to weighted least squares for the multiple fundamental case, and presents an application in signal processing.
Key words and phrases: Asymptotic variance, harmonic regression, signal processing, sound analysis, time-varying parameters, weighted least squares estimates.