Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Semi-spontaneous Dutch Speech Dataset for Speech Enhancement and Speech Recognition

Dimme de Groot, Yuanyuan Zhang, Jorge Martinez, Odette Scharenborg

TL;DR

It is found that five out of the eight ASR models have WERs lower than 22% on DRES, despite the challenging conditions, and no positive effect of modern single-channel SE on ASR performance is found.

Abstract

We present DRES: a 1.5-hour Dutch realistic elicited (semi-spontaneous) speech dataset from 80 speakers recorded in noisy, public indoor environments. DRES was designed as a test set for the evaluation of state-of-the-art (SOTA) automatic speech recognition (ASR) and speech enhancement (SE) models in a real-world scenario: a person speaking in a public indoor space with background talkers and noise. The speech was recorded with a four-channel linear microphone array. In this work we evaluate the speech quality of five well-known single-channel SE algorithms and the recognition performance of eight SOTA off-the-shelf ASR models before and after applying SE on the speech of DRES. We found that five out of the eight ASR models have WERs lower than 22% on DRES, despite the challenging conditions. In contrast to recent work, we did not find a positive effect of modern single-channel SE on ASR performance, emphasizing the importance of evaluating in realistic conditions.

A Semi-spontaneous Dutch Speech Dataset for Speech Enhancement and Speech Recognition

TL;DR

It is found that five out of the eight ASR models have WERs lower than 22% on DRES, despite the challenging conditions, and no positive effect of modern single-channel SE on ASR performance is found.

Abstract

We present DRES: a 1.5-hour Dutch realistic elicited (semi-spontaneous) speech dataset from 80 speakers recorded in noisy, public indoor environments. DRES was designed as a test set for the evaluation of state-of-the-art (SOTA) automatic speech recognition (ASR) and speech enhancement (SE) models in a real-world scenario: a person speaking in a public indoor space with background talkers and noise. The speech was recorded with a four-channel linear microphone array. In this work we evaluate the speech quality of five well-known single-channel SE algorithms and the recognition performance of eight SOTA off-the-shelf ASR models before and after applying SE on the speech of DRES. We found that five out of the eight ASR models have WERs lower than 22% on DRES, despite the challenging conditions. In contrast to recent work, we did not find a positive effect of modern single-channel SE on ASR performance, emphasizing the importance of evaluating in realistic conditions.
Paper Structure (17 sections, 3 figures, 2 tables)

This paper contains 17 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. The DRES corpus
  3. Corpus design
  4. The speech recordings
  5. Participants
  6. Transcriptions
  7. The quality of the speech data
  8. Experiments
  9. Speech enhancement algorithms
  10. Automatic speech recognition models
  11. Evaluation metrics
  12. Results
  13. Speech quality
  14. ASR performance
  15. Discussion and conclusions
  16. ...and 2 more sections

Figures (3)

  • Figure 1: An example picture-card (a) and prompt-card (b). (c): The four-channel microphone array used during the recording.
  • Figure 2: The distribution of the mean-opinion-score (MOS) for the recordings at the four recording locations, estimated using DNSMOS. Higher values imply a better quality. 'Overall' represents the combined results over all locations.
  • Figure 3: Left: Estimated distribution of DNSMOS for the baseline signal and each of the SE algorithms. Higher values imply a better overall quality. Right: DNSMOS score improvements with respect to the baseline for each of the SE algorithms.