A user controlled system for the generation of melodies applying case based reasoning

Abstract

The automatic generation of music is an emergent field of research that has attracted a wide number of investigators. Many systems allow a collaboration between human and machine to generate valuable music. Among the different approaches developed in the state of the art, the present research is focused on an intelligent system that generates melodies through a mechanical device guided by the user. The system is able to learn from previous compositions created by the users to improve future results. A Case-Based Reasoning architecture was developed with a Markov model to obtain the probabilities of a given note following the last note incorporated in the melody. This probability also depends on the mechanical device connected to the system that can be used at any moment to control the pitches and the duration of the musical notes. As a result of the collaboration between machine and user, we obtain a melody that will be rated and, according to the rating, incorporated into the memory of the system for future use. Several experiments were developed to analyze the quality of the system and the melodies created. The results of the experiments reveal that the proposed system is able to generate music adapted and controlled by the users.