Sound timbre interpolation based on physical modeling

Our goal is to develop sound synthesis technology that users can synthesize arbitrary sound timbre, including musical instrument sounds, natural sounds, and their interpolation/extrapolation on demand. For this purpose, we investigated sound interpolation based on physical modeling. A sound-synthesis model composed of an exciter, a one-dimensional vibrator, and a two-dimensional resonator is used, and smooth timbre conversion by parameter control is examined. Piano and guitar sounds are simulated using this model, and interpolation between piano and guitar tones is investigated. The strategy for parameter control is proposed, and subjective tests were performed to evaluate the algorithm. A multidimensional scaling (MDS) technique is used, and perceptual characteristics are discussed. One of the axes of the timbre space is interpreted as spectral energy distribution, so the spectral centroid is used as a reference to adjust parameters for synthesis. By considering the centroids, smoothly interpolating timbre is achieved. These results suggest the possibility of developing a morphing system using a physical model.