Dynamic Sliding Window for Realtime Denoising Networks

Realtime speech denoising has been long studied. Almost all existing methods process the incoming data stream using a sliding window of fixed-size. Yet, we show that the use of fixed-size sliding window may lead to an accumulating lag, especially in presence of other background computing processes that may occupy CPU resources. In response, we propose a new sliding window strategy and a lightweight neural network to leverage it. Our experiments show that the proposed approach achieves denoising quality on a par with the stateof-the-art realtime denoising models. More importantly, our approach is faster, maintaining a stable realtime performance even when the available computing power fluctuates.

1. INTRODUCTION

Realtime speech denoising is a highly demanded audio processing task—perhaps more demanded than ever—as our world is still shadowed by the COVID pandemic and the online meeting is becoming a "new normal" of our daily social life. Most recent, state-of-the-art realtime denoising techniques are all based on neural networks [1, 2, 3, 4, 5, 6, 7, 8]. They seek novel network structures to achieve plausible denoising quality while retaining network simplicity to reduce processing time.