An LSTM-Based Chord Generation System Using Chroma Histogram Representations

TL;DR

This paper introduces an LSTM-based system for chord generation from monophonic melodies using chroma histogram representations, offering harmonic richness and real-time applicability, though with some limitations in long-term coherence.

Contribution

It presents a novel chord generation approach leveraging chroma histograms with LSTM, emphasizing harmonic richness and real-time suitability over traditional chord label methods.

Findings

Suitable for real-time chord generation

Demonstrates diatonic and cadential relationships

Does not achieve state-of-the-art long-term coherence

Abstract

This paper proposes a system for chord generation to monophonic symbolic melodies using an LSTM-based model trained on chroma histogram representations of chords. Chroma representations promise more harmonically rich generation than chord label-based approaches, whilst maintaining a small number of dimensions in the dataset. This system is shown to be suitable for limited real-time use. While it does not meet the state-of-the-art for coherent long-term generation, it does show diatonic generation with cadential chord relationships. The need for further study into chroma histograms as an extracted feature in chord generation tasks is highlighted.