Automatic Melody Composition and Evolution: A Cognitive-Based Approach

Music composition is a highly interdisciplinary process. To understand it deeply, a number of approaches have been used from different fields, such as musicology, music theory, music cognition and philosophy. During recent decades, numerous techniques based on Artificial Intelligence (AI) have been proposed. In particular, many AI systems focus on automatic melodic composition. Most of these systems try to generate melodies enjoyable by a human, but they completely ignore the way in which humans actually compose. Humans create music by exploiting a mixed top-down bottom-up approach, characterised by high-level cognition processes and rules.
In this paper, we propose a memetic model for music composition, which considers both psychological and social levels. The former level analyses the actual cognitive mechanisms and procedures involved while composing music: namely, museme network, compositional grammar and evaluation module. The social level puts the figure of the composer into perspective within her musical environment. The introduced memetic model is encoded in a two-step algorithm. Firstly, a top-down approach is used for defining the overall structure of a melody. Secondly, the given structure is filled with musical content, following a bottom-up strategy, that fosters emergent behaviour. The proposed algorithm is the first system we are aware of which can evolve its own compositional style. Stylistic change is achieved by modifying grammar rules and the museme network. Finally, the paper provides an analysis of generated melodies.