Publication Details

Building and Evaluation of a Real Room Impulse Response Dataset

SZŐKE Igor, SKÁCEL Miroslav, MOŠNER Ladislav, PALIESEK Jakub and ČERNOCKÝ Jan. Building and Evaluation of a Real Room Impulse Response Dataset. IEEE Journal of Selected Topics in Signal Processing, vol. 13, no. 4, 2019, pp. 863-876. ISSN 1932-4553. Available from: https://ieeexplore.ieee.org/document/8717722
Czech title
Tvorba a vyhodnocení databáze reálných impulzních odezev místností
Type
journal article
Language
english
Authors
Szőke Igor, Ing., Ph.D. (DCGM FIT BUT)
Skácel Miroslav, Ing. (DCGM FIT BUT)
Mošner Ladislav, Ing. (DCGM FIT BUT)
Paliesek Jakub, Ing. (FIT BUT)
Černocký Jan, prof. Dr. Ing. (DCGM FIT BUT)
URL
Keywords

far-field, automatic speech recognition, room impulse response, reverberation, SineSweep, Maximum Length Sequence, noise, deep neural network, Kaldi, AMI

Abstract

This paper presents BUT ReverbDB - a dataset of real room impulse responses (RIR), background noises and re-transmitted speech data. The retransmitted data includes LibriSpeech test-clean, 2000 HUB5 English evaluation and part of 2010 NIST Speaker Recognition Evaluation datasets. We provide a detailed description of RIR collection (hardware, software, post-processing) that can serve as a "cook-book" for similar efforts. We also validate BUT ReverbDB in two sets of automatic speech recognition (ASR) experiments and draw conclusions for augmenting ASR training data with real and artificially generated RIRs. We show that a limited number of real RIRs, carefully selected to match the target environment, provide results comparable to a large number of artificially generated RIRs, and that both sets can be combined to achieve the best ASR results. The dataset is distributed for free under a non-restrictive license and it currently contains data from 8 rooms, which is growing. The distribution package also contains a Kaldi-based recipe for augmenting publicly available AMI close-talk meeting data and test the results on an AMI single distant microphone set, allowing it to reproduce our experiments.

Published
2019
Pages
863-876
Journal
IEEE Journal of Selected Topics in Signal Processing, vol. 13, no. 4, ISSN 1932-4553
Publisher
Institute of Electrical and Electronics Engineers
DOI
UT WoS
000477715300008
EID Scopus
BibTeX
@ARTICLE{FITPUB11982,
   author = "Igor Sz\H{o}ke and Miroslav Sk\'{a}cel and Ladislav Mo\v{s}ner and Jakub Paliesek and Jan \v{C}ernock\'{y}",
   title = "Building and Evaluation of a Real Room Impulse Response Dataset",
   pages = "863--876",
   journal = "IEEE Journal of Selected Topics in Signal Processing",
   volume = 13,
   number = 4,
   year = 2019,
   ISSN = "1932-4553",
   doi = "10.1109/JSTSP.2019.2917582",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/11982"
}
Back to top