Synthesizer sound programming requires adjusting a large number of parameters, making it a complex task, especially for beginners. To address this issue, prior research has proposed support systems that suggest parameter adjustments based on large preset datasets labeled with perceptual sound descriptors such as “Soft” or “Bright.” However, these systems are often specific to synthesizers with rich preset libraries, making it difficult to apply them across different devices. This study aims to construct a generalized synthesizer programming support system that can be applied to multiple synthesizers by introducing acoustic features as a common representation across devices. In the proposed system, the user is guided to adjust parameters that strongly affect acoustic features corresponding to a target descriptor label, thereby supporting intuitive creation of original sounds. As a preliminary study, we analyzed labeled presets from a commercial synthesizer and extracted various acoustic features. Using random forest classifiers, we calculated the feature importance for each label classification task and evaluated the relationships between labels and acoustic features. The results revealed clear associations between certain labels and acoustic features, suggesting the potential of leveraging these findings to mediate between labels and parameters. This study provides fundamental insights for synthesizer-independent sound design systems.
Kobata et al. (Wed,) studied this question.