Lightweight Online Noise Reduction on Embedded Devices using Hierarchical Recurrent Neural Networks

TL;DR

This paper introduces a lightweight, hierarchical recurrent neural network-based noise reduction method optimized for embedded devices like hearing aids, achieving high noise suppression with minimal parameters and computational cost.

Contribution

The authors develop a hierarchical RNN architecture that drastically reduces model size and FLOPs, enabling real-time noise reduction on embedded devices without sacrificing performance.

Findings

Model with only 5k parameters performs well on unseen noise.

Achieves noise reduction comparable to larger models.

Suitable for low-latency, online processing on embedded hardware.

Abstract

Deep-learning based noise reduction algorithms have proven their success especially for non-stationary noises, which makes it desirable to also use them for embedded devices like hearing aids (HAs). This, however, is currently not possible with state-of-the-art methods. They either require a lot of parameters and computational power and thus are only feasible using modern CPUs. Or they are not suitable for online processing, which requires constraints like low-latency by the filter bank and the algorithm itself. In this work, we propose a mask-based noise reduction approach. Using hierarchical recurrent neural networks, we are able to drastically reduce the number of neurons per layer while including temporal context via hierarchical connections. This allows us to optimize our model towards a minimum number of parameters and floating-point operations (FLOPs), while preserving noise reduction quality compared to previous work. Our smallest network contains only 5k parameters, which makes this algorithm applicable on embedded devices. We evaluate our model on a mixture of EUROM and a real-world noise database and report objective metrics on unseen noise.