This paper presents a simple and easy way to obtain good estimates of the frequencies of spectral peaks in voiced sounds. The instantaneous frequency throughout the spectrum is calculated using an analytically derived window. Peaks are detected by examining the difference between the frequencies of spectral lines and their corresponding instantaneous frequencies. The instantaneous frequencies near peaks are used as estimates of their actual frequencies. The corresponding amplitude values are calculated from the Gabor transform at the estimated frequencies. The performance of the estimation of the frequencies using analytically derived windows has been verified using synthetic signals and musical sounds. This method gives a better estimate of the instantaneous frequency than commonly used methods and the resynthesis of sound from the additive parameters gives good results.

Download

This paper presents a model of the envelope of the additive parameters of isolated musical sounds, along with a new method for the estimation of the important envelope splitpoint times. The model consists of start, attack, sustain, release, and end segments with variable split-point amplitude and time. The estimation of the times is done using smoothed derivatives of the envelopes. The estimated split-point values can be used together with a curve-form model introduced in this paper in the analysis/synthesis of musical sounds. The envelope model can recreate noise-less musical sounds with good fidelity, and the method for the estimation of the envelope times performs significantly better than the classical percentagebased method.

Download