Spatial Diffuseness Features for DNN-Based Speech Recognition in Noisy and Reverberant Environments

TL;DR

This paper introduces a real-time spatial diffuseness feature for DNN-based speech recognition that enhances accuracy in noisy, reverberant environments by representing diffuse noise levels without needing direction of arrival estimation.

Contribution

The paper presents a novel diffuseness feature that improves speech recognition accuracy in challenging environments without requiring complex spatial information estimation.

Findings

Reduced word error rate on REVERB challenge corpus

Outperforms logmelspec and spectral subtraction features

Effective in real-time noisy and reverberant conditions

Abstract

We propose a spatial diffuseness feature for deep neural network (DNN)-based automatic speech recognition to improve recognition accuracy in reverberant and noisy environments. The feature is computed in real-time from multiple microphone signals without requiring knowledge or estimation of the direction of arrival, and represents the relative amount of diffuse noise in each time and frequency bin. It is shown that using the diffuseness feature as an additional input to a DNN-based acoustic model leads to a reduced word error rate for the REVERB challenge corpus, both compared to logmelspec features extracted from noisy signals, and features enhanced by spectral subtraction.