Modeling Piano Tones with Group Synthesis
Kathy Lee, Andrew Horner · Rare & Special e-Zone (The Hong Kong University of Science and Technology) · 1999
Most synthesizers and sound cards use sampling to synthesize the piano. Sampling requires a lot of memory and does not allow much flexibility, unlike methods such as wavetable synthesis. However, wavetable synthesis usually assumes that the original tone is harmonic, and the piano's stretched partials make it difficult to model. It is shown how to simulate piano tones using group synthesis, a special case of wavetable synthesis. The synthesis model uses nearly contiguous groups selected by a genetic algorithm, and independent frequency deviations for each group to simulate partial stretching. The model reduces the amount of data by about 50% compared to additive synthesis, while preserving the perceptual identity of the sound. Listening test results indicate that the synthesized piano tones are perceptually similar to the originals.