Learning Disentangled Representations of Timbre and Pitch for Musical Instrument Sounds Using Gaussian Mixture Variational Autoencoders

TL;DR

This paper introduces a Gaussian mixture variational autoencoder framework that learns to disentangle timbre and pitch in musical instrument sounds, enabling controllable synthesis and transfer of instrument characteristics.

Contribution

The work presents a novel disentanglement approach using Gaussian mixture VAEs with separate encoders for timbre and pitch, facilitating realistic sound synthesis and transfer.

Findings

High accuracy in instrument and pitch classification on synthesized sounds

Effective disentanglement of timbre and pitch demonstrated through visualization and analysis

Successful timbre transfer between instruments using a single autoencoder architecture

Abstract

In this paper, we learn disentangled representations of timbre and pitch for musical instrument sounds. We adapt a framework based on variational autoencoders with Gaussian mixture latent distributions. Specifically, we use two separate encoders to learn distinct latent spaces for timbre and pitch, which form Gaussian mixture components representing instrument identity and pitch, respectively. For reconstruction, latent variables of timbre and pitch are sampled from corresponding mixture components, and are concatenated as the input to a decoder. We show the model efficacy by latent space visualization, and a quantitative analysis indicates the discriminability of these spaces, even with a limited number of instrument labels for training. The model allows for controllable synthesis of selected instrument sounds by sampling from the latent spaces. To evaluate this, we trained instrument and pitch classifiers using original labeled data. These classifiers achieve high accuracy when tested on our synthesized sounds, which verifies the model performance of controllable realistic timbre and pitch synthesis. Our model also enables timbre transfer between multiple instruments, with a single autoencoder architecture, which is evaluated by measuring the shift in posterior of instrument classification. Our in depth evaluation confirms the model ability to successfully disentangle timbre and pitch.