Estimating sinusoidal parameters of musical tones based on global waveform fitting

A novel approach to the analysis of musical tones using a quadratic polynomial phase sinusoids plus residual data model is presented. Taking advantage of the fact that musical signals are usually analyzed off-line, the proposed approach estimates the sinusoidal parameters of a musical tone such that its waveform fits the data model in a least square sense, and has obvious advantages over the existing sinusoidal model-based music analysis/synthesis algorithms in terms of data reduction and computational complexity. Sound examples are included to demonstrate the effectiveness of the proposed approach.