Generation of Pop-Rock Chord Sequences Using Genetic Algorithms and Variable Neighborhood Search

摘要

This work proposes a utility function that measures: 1) the vertical relation between notes in a melody and chords in a sequence, and 2) the horizontal relation among chords. This utility function is embedded in a procedure that combines a Genetic Algorithm (GA) with a Variable Neighborhood Search (VNS) to automatically generate style-based chord sequences. The two-step algorithm is tested in ten popular songs, achieving accompaniments that match closely those of the original versions.