Splicing-Inspired Recognition and Composition of Musical Collectives Styles

Computer music is an emerging area for the application of computational techniques inspired by information processing in Nature. A challenging task in this area is the automatic recognition of musical styles. The style of a musician is the result of the combination of several factors such as experience, personality, preferences. In the last years, several works have been proposed for the recognition of styles for soloists performers, where the improvisation often plays an important role. The evolution of this problem, that is the recognition of multiple performers' style that collaborate over time to perform, record or compose music, know as Musical collective, presents many more difficulties, due to the simultaneous presence of various performers, mutually conditionable. In this paper, we propose a new approach for both recognition and automatic composition of styles for musical collectives. Specifically, our system exploits a machine learning recognizer, based on one-class support vector machines and neural networks for style recognition, and a splicing composer, for music composition (in the style of the whole collective). To assess the effectiveness of our system we performed several tests using transcriptions of popular jazz bands. With regard to the recognition, we show that our classifier is able to achieve an accuracy of 97.7%. With regard to the composition, we measured the quality of the generated compositions by collecting subjective perceptions from domain experts.