Self-Attention for Audio Super-Resolution

TL;DR

This paper introduces a novel audio super-resolution network that combines convolutional layers with self-attention mechanisms, capturing long-range dependencies more effectively and enabling faster training.

Contribution

It proposes a new architecture integrating self-attention with convolutional networks and introduces AFiLM for improved modulation in audio super-resolution.

Findings

Outperforms existing methods on standard benchmarks

Enables more parallelization and faster training

Effectively models long-range dependencies in audio sequences

Abstract

Convolutions operate only locally, thus failing to model global interactions. Self-attention is, however, able to learn representations that capture long-range dependencies in sequences. We propose a network architecture for audio super-resolution that combines convolution and self-attention. Attention-based Feature-Wise Linear Modulation (AFiLM) uses self-attention mechanism instead of recurrent neural networks to modulate the activations of the convolutional model. Extensive experiments show that our model outperforms existing approaches on standard benchmarks. Moreover, it allows for more parallelization resulting in significantly faster training.