Parameter Estimation and Tracking for Time-varying Sinusoids

Parametric modeling permits an efficient representation of audio signals and is increasingly utilized for very low bit rate coding applications. Such systems are based on a decomposition of the audio signal into components that are described by appropriate source models and represented by model parameters. Commonly used components types are sinusoidal trajectories, harmonic tones, transients, and noise. Proper estimation and tracking of sinusoids in case of vibrato or portamento is vital, yet difficult due to the uncertainty principle for time-frequency resolution. This paper presents a reliable solution to this problem that was successfully demonstrated in an encoder for MPEG-4 parametric audio coding “Harmonic and Individual Lines plus Noise” (HILN).