Emotion-driven Piano Music Generation via Two-stage Disentanglement and Functional Representation

TL;DR

This paper presents a two-stage framework for emotion-driven piano music generation that disentangles valence and arousal, utilizing a novel functional representation to better capture emotional features and enable controllable music synthesis.

Contribution

It introduces a two-stage disentanglement approach for valence and arousal in piano music generation, along with a new functional symbolic music representation for improved emotional modeling.

Findings

Effective modeling of emotional valence and arousal validated by experiments.

Functional representation captures major-minor tonality and note interactions.

Framework enables controllable emotion-driven music generation.

Abstract

Managing the emotional aspect remains a challenge in automatic music generation. Prior works aim to learn various emotions at once, leading to inadequate modeling. This paper explores the disentanglement of emotions in piano performance generation through a two-stage framework. The first stage focuses on valence modeling of lead sheet, and the second stage addresses arousal modeling by introducing performance-level attributes. To further capture features that shape valence, an aspect less explored by previous approaches, we introduce a novel functional representation of symbolic music. This representation aims to capture the emotional impact of major-minor tonality, as well as the interactions among notes, chords, and key signatures. Objective and subjective experiments validate the effectiveness of our framework in both emotional valence and arousal modeling. We further leverage our framework in a novel application of emotional controls, showing a broad potential in emotion-driven music generation.