Using Monte Carlo dropout for non-stationary noise reduction from speech

TL;DR

This paper introduces a Bayesian approach using Monte Carlo dropout to improve speech enhancement neural networks, enabling better generalization to unseen noise conditions and dynamic model selection based on estimated model precision.

Contribution

The work applies MC dropout for Bayesian uncertainty estimation in speech enhancement, allowing adaptive model selection and improved performance on unseen noise types.

Findings

MC dropout improves speech enhancement in unseen noise conditions.

Dynamic model selection based on model precision enhances robustness.

Proposed method outperforms traditional models in non-stationary noise environments.

Abstract

In this work, we propose the use of dropout as a Bayesian estimator for increasing the generalizability of a deep neural network (DNN) for speech enhancement. By using Monte Carlo (MC) dropout, we show that the DNN performs better enhancement in unseen noise and SNR conditions. The DNN is trained on speech corrupted with Factory2, M109, Babble, Leopard and Volvo noises at SNRs of 0, 5 and 10 dB. Speech samples are obtained from the TIMIT database and noises from NOISEX-92. In another experiment, we train five DNN models separately on speech corrupted with Factory2, M109, Babble, Leopard and Volvo noises, at 0, 5 and 10 dB SNRs. The model precision (estimated using MC dropout) is used as a proxy for squared error to dynamically select the best of the DNN models based on their performance on each frame of test data. We propose an algorithm with a threshold on the model precision to switch between classifier based model selection scheme and model precision based selection scheme. Testing is done on speech corrupted with unseen noises White, Pink and Factory1 and all five seen noises.