Generalized overlap-add sinusoidal modeling applied to quasi-harmonic tone synthesis

Abstract

Analysis-by-synthesis/overlap-add (ABS/OLA) sinusoidal modeling has been successfully demonstrated as an accurate, flexible, and computationally tractable representation for the purposes of speech modification and harmonic tone synthesis; however, the model formulation used to synthesize these signals does not take full advantage of the structure of quasi-harmonic music signals. This paper describes a generalized overlap-add sinusoidal model formulation that accounts for the time-frequency behavior of quasi-harmonic tones and which reduces to the previous formulation as a special case.<>