Normalized Features for Improving the Generalization of DNN Based Speech Enhancement

TL;DR

This paper introduces normalized features based on SNR estimates to improve the generalization of deep neural network speech enhancement models, leading to better noise suppression in unseen conditions.

Contribution

It proposes a novel feature set combining a priori and a posteriori SNRs to enhance DNN-based speech enhancement and improve generalization to unknown noise types.

Findings

Significant PESQ and STOI improvements in unseen noise conditions.

Enhanced generalization of DNN models with proposed features.

Listening tests confirm better speech quality and intelligibility.

Abstract

Enhancing noisy speech is an important task to restore its quality and to improve its intelligibility. In traditional non-machine-learning (ML) based approaches the parameters required for noise reduction are estimated blindly from the noisy observation while the actual filter functions are derived analytically based on statistical assumptions. Even though such approaches generalize well to many different acoustic conditions, the noise suppression capability in transient noises is low. To amend this shortcoming, machine-learning (ML) methods such as deep learning have been employed for speech enhancement. However, due to their data-driven nature, the generalization of ML based approaches to unknown noise types is still discussed. To improve the generalization of ML based algorithms and to enhance the noise suppression of non-ML based methods, we propose a combination of both approaches. For this, we employ the a priori signal-to-noise ratio (SNR) and the a posteriori SNR estimated as input features in a deep neural network (DNN) based enhancement scheme. We show that this approach allows ML based speech estimators to generalize quickly to unknown noise types even if only few noise conditions have been seen during training. Further, the proposed features outperform a competing approach where an estimate of the noise power spectral density is appended to the noisy spectra. Instrumental measures such as Perceptual Evaluation of Speech Quality (PESQ) and short-time objective intelligibility (STOI) indicate strong improvements in unseen conditions when the proposed features are used. Listening experiments confirm the improved generalization of our proposed combination.