A Sound Analysis and Synthesis System for Generating an Instrumental Piri Song

This paper proposes spectral modeling of the piri instrument based on its acoustic characteristics. Since the harmonics of piri sounds are evenly distributed to the high frequency band (up to 20 kHz), spectral modeling synthesis is suitable for generating piri sounds. In this study, sinusoids of the piri are only considered with formant models (or resonant models) to reduce the number of synthetic parameters using cepstral analysis. In addition, noise components are recreated by exploiting a line-segmentation approximation. Experimental results indicated that single notes of the piri and a synthesized instrumental piri song have similar spectra to originals.