Exploring the robustness of features and enhancement on speech recognition systems in highly-reverberant real environments

TL;DR

This study assesses the robustness of speech recognition systems in highly reverberant environments, comparing feature extraction and enhancement techniques under different training conditions to improve performance.

Contribution

It evaluates the effectiveness of various feature and enhancement combinations for DNN-HMM speech recognition in reverberant settings, highlighting the impact of training conditions.

Findings

WPE and LNFB reduce WERs more with reverberant training.

LNFB outperforms MelFB under clean training conditions.

Different methods complement each other depending on environment and training.

Abstract

This paper evaluates the robustness of a DNN-HMM-based speech recognition system in highly-reverberant real environments using the HRRE database. The performance of locally-normalized filter bank (LNFB) and Mel filter bank (MelFB) features in combination with Non-negative Matrix Factorization (NMF), Suppression of Slowly-varying components and the Falling edge (SSF) and Weighted Prediction Error (WPE) enhancement methods are discussed and evaluated. Two training conditions were considered: clean and reverberated (Reverb). With Reverb training the use of WPE and LNFB provides WERs that are 3% and 20% lower in average than SSF and NMF, respectively. WPE and MelFB provides WERs that are 11% and 24% lower in average than SSF and NMF, respectively. With clean training, which represents a significant mismatch between testing and training conditions, LNFB features clearly outperform MelFB features. The results show that different types of training, parametrization, and enhancement techniques may work better for a specific combination of speaker-microphone distance and reverberation time. This suggests that there could be some degree of complementarity between systems trained with different enhancement and parametrization methods.