Blind Reverberation Time Estimation in Dynamic Acoustic Conditions

TL;DR

This paper introduces a new data generation method and demonstrates improved deep neural network performance for real-time reverberation time estimation in changing acoustic environments.

Contribution

It presents a novel data synthesis approach and adapts existing neural networks to better track dynamic reverberation conditions.

Findings

Enhanced neural network accuracy in dynamic environments

Significant improvement over static-condition models

Effective real-time reverberation time tracking

Abstract

The estimation of reverberation time from real-world signals plays a central role in a wide range of applications. In many scenarios, acoustic conditions change over time which in turn requires the estimate to be updated continuously. Previously proposed methods involving deep neural networks were mostly designed and tested under the assumption of static acoustic conditions. In this work, we show that these approaches can perform poorly in dynamically evolving acoustic environments. Motivated by a recent trend towards data-centric approaches in machine learning, we propose a novel way of generating training data and demonstrate, using an existing deep neural network architecture, the considerable improvement in the ability to follow temporal changes in reverberation time.