ORPHEU - A PERFORMANCE-ORIENTED SYSTEM FOR MUSICAL ACCOMPANIMENT: AN APPROACH USING THE MARKOV DECISION PROCESS
Pedro de Botelho Marcos, Leonardo Ramos Emmendörfer · 2012
Nowadays, applications have an increasingly degree of interactivity between the computer and the users. A great part of them are multimedia applications, and among them are the computer music system. It is the eld that combines musical techniques with computational algorithms. One interesting application of computer music is the Computer-Aided Algorithmic Composition. The composition may be an entire song or just a part of it. This paper presents the Orpheu architecture, which uses the Markov decision process to predict, in real time, an accompaniment for a solo. Also, a learning mechanism was developed to allow the system to learn how musician performs his solos. To validate the system experiments were performed and measured the correctness of the generated accompaniment. In a real scenario the system achieves an average hit rate about 90%.