The Effect of Perceptual Metrics on Music Representation Learning for Genre Classification

TL;DR

This paper shows that using perceptual metrics as loss functions in autoencoder training enhances music genre classification by capturing meaningful features, leading to better generalization to new signals.

Contribution

It introduces a novel approach of employing perceptual metrics as loss functions in autoencoders for music representation learning, improving genre classification performance.

Findings

Autoencoder features trained with perceptual losses outperform direct metric-based methods.

Perceptual loss functions improve generalization to unseen music signals.

Using perceptual metrics as loss functions captures meaningful, human-aligned features.

Abstract

The subjective quality of natural signals can be approximated with objective perceptual metrics. Designed to approximate the perceptual behaviour of human observers, perceptual metrics often reflect structures found in natural signals and neurological pathways. Models trained with perceptual metrics as loss functions can capture perceptually meaningful features from the structures held within these metrics. We demonstrate that using features extracted from autoencoders trained with perceptual losses can improve performance on music understanding tasks, i.e. genre classification, over using these metrics directly as distances when learning a classifier. This result suggests improved generalisation to novel signals when using perceptual metrics as loss functions for representation learning.