Audio Super Resolution using Neural Networks

TL;DR

This paper presents a neural network-based audio super-resolution method that enhances low-quality audio signals by predicting missing samples, outperforming baselines in speech and music benchmarks for various upscaling ratios.

Contribution

Introduces a simple deep convolutional neural network approach for audio super-resolution that does not rely on specialized audio processing techniques.

Findings

Outperforms baseline methods on speech and music benchmarks

Effective at 2x, 4x, and 6x upscaling ratios

Demonstrates practical applications in telephony and TTS

Abstract

We introduce a new audio processing technique that increases the sampling rate of signals such as speech or music using deep convolutional neural networks. Our model is trained on pairs of low and high-quality audio examples; at test-time, it predicts missing samples within a low-resolution signal in an interpolation process similar to image super-resolution. Our method is simple and does not involve specialized audio processing techniques; in our experiments, it outperforms baselines on standard speech and music benchmarks at upscaling ratios of 2x, 4x, and 6x. The method has practical applications in telephony, compression, and text-to-speech generation; it demonstrates the effectiveness of feed-forward convolutional architectures on an audio generation task.