Pitch Gestures in Generative Modeling of Music

Generative models of music are in need of performance and gesture additions, i.e. inclusions of subtle temporal and dynamic alterations, and gestures so as to render the music musical. While much of the research regarding music generation is based on music theory, the work presented here is based on the temporal perception, which is divided into three parts, the immediate (subchunk), the short-term memory (chunk), and the superchunk. By review of the relevant temporal perception literature, the necessary performance elements to add in the metrical generative model, related to the chunk memory, are obtained. In particular, the pitch gestures are modeled as rising, falling, or as arches with positive or negative peaks.