Iterative method for Harmonic and Exponentially Damped Sinusoidal model
Abstract
We present two evolutions of the well known Exponentially Damped Sinusoidal model, named HDS (Harmonic Exponentially Damped Sinusoidal) model for single pitch signals and HDSM for multi-pitch signals modeling. Additionally, we propose to use an iterative high-accuracy algorithm to estimate the model parameters. Compared with full or fast implementations of high-resolution methods (Matrix Pencil, ESPRIT, Kung’s algorithm,...), this approach has a lower computational complexity and is well adapted to the Harmonic Damped Sinusoidal models since it takes into account the harmonic relations between the angular- frequencies. In the context of low bit-rate audio coding, we show the validity of this approach on several audio signals.