Optimal LS IIR filter design for music analysis/synthesis

Addresses the design of fixed, low-order infinite impulse response (IIR) filters for modeling the perceptually significant features of the spectra of string instrument bodies. The problem is stated mathematically, and the design methodologies compared here are reviewed. The experimental results are presented. The experimental set-up, data acquisition and data preprocessing are described. The spectra of the IIR filters designed using the least squares modified Yule-Walker equation, the Steiglitz-Mcbride method, and Hankel norm methods are compared. These spectra are also compared to the spectrum obtained by using an equivalent-order autoregressive model computed using the autocorrelation method. Results are summarized and ongoing work is discussed