Machine Learning Based Fuzzy Logic Approach for Auto Harmonization of Melodies in C Major

Abstract

Melody harmonization has been a matter of interest for musicians over the past centuries. Numerous efforts have been made in harmonically arranging music both manually and automatically. As harmonization is evolved from chord progressions, sincere efforts have been made to emulate Bach’s compositional methods to create impressive chord patterns using algorithmic approaches. Different artificial intelligence techniques involving supervised learning have also been subsequently applied. In this paper, we propose a fuzzy logic approach for automatically harmonizing melodies in C major scale. The model is built on the basis of data gathered from musical corpora over the ages. The melody is harmonized utilizing a fuzzy reasoning-based automated harmonization technique. Through the fuzzy reasoning applied, typical chord progressions resembling legendary manual compositions are obtained. We have used basic chords avoiding complex textures so that the tonality of the music is retained. The performance of the system is evaluated comprehensively. The system shows competency on comparative analysis with manual harmonization done by musicians on the same melody.