NONSTATIONARY SINUSOIDAL MODELING WITH EFFICIENT ESTIMATION OF LINEAR FREQUENCY CHIRP PARAMETERS

Stanford University,
Center for Computer Research in Music and Acoustics,
Stanford, CA 94305-8180 USA

Abstract:

We present an enhanced sinusoidal modeling system that efficiently parameterizes spectral peaks by linear frequency chirp rate, in addition to the standard parameters of amplitude, center frequency, and phase. Similar to a conventional sinusoidal modeling systems, the current system operates in a frame-by-frame manner, but also efficiently obtains a chirp parameter estimate for each peak in a given frame. We demonstrate an application to voice coding where our model captures most of the significant signal information in a spectrogram operating at double the current system's frame rate.