GraFPrint: A GNN-Based Approach for Audio Identification

TL;DR

GraFPrint is a novel GNN-based framework that creates robust audio fingerprints by leveraging graph structures and self-supervised learning, showing superior scalability and resilience to distortions in large-scale datasets.

Contribution

It introduces GraFPrint, a new GNN-based audio identification method that constructs k-NN graphs and uses contrastive training for improved robustness and scalability.

Findings

Outperforms existing methods on large-scale datasets

Resilient to ambient distortions due to contrastive training

Lightweight and scalable for real-world applications

Abstract

This paper introduces GraFPrint, an audio identification framework that leverages the structural learning capabilities of Graph Neural Networks (GNNs) to create robust audio fingerprints. Our method constructs a k-nearest neighbor (k-NN) graph from time-frequency representations and applies max-relative graph convolutions to encode local and global information. The network is trained using a self-supervised contrastive approach, which enhances resilience to ambient distortions by optimizing feature representation. GraFPrint demonstrates superior performance on large-scale datasets at various levels of granularity, proving to be both lightweight and scalable, making it suitable for real-world applications with extensive reference databases.