Weighted Estimation of Harmonic Components in a Musical Sound Signal (Q4544837)

The present paper shows that regression models can be a useful tool for the analysis of music signals. The model under consideration is \[ y(t)=\sum_{k=1}^K\{A_k\cos(\omega_kt)+ B_k\sin(\omega_kt)\}+\varepsilon(t),\quad t=1,\dots,T, \] where \(\{\varepsilon(t)\}\) is a stationary stochastic process. Weighted least squares estimates for the model parameters are studied, an asymptotic variance expression is obtained and consistency of the weighted least squares estimates is proved. Particularly, the author uses the asymptotic variance expressions to develop standard errors and to construct confidence intervals for regression parameters. Finally, he presents possible applications of the proposed statistical technique to analyse musical signals (for instance, clarinet and guitar sounds).