Result Details
BUT QUESST 2015 System Description
Szőke Igor, Ing., Ph.D., DCGM (FIT)
All our systems are based on Dynamic Time Warping (DTW). These systems use bottle-neck features (BN) as input. The bottle-neck feature extractors were trained on Global Phone Czech, Portuguese, Russian and Spanish languages, so our approach is in low-resource category. We also aimed onT1/T2/T3 types of query search for late submission sys-tems. System calibration and fusion were based on binarylogistic regression.
query-by-example, DTW, dynamic time warping, spoken term detection, keyword spotting, bottle-neck features
This paper describes system for query-by-example spoken term detection built for QUESST 2015 evaluation. The system uses dynamic time warping to detect spoken term utilizing bottle-neck features.
@inproceedings{BUT120391,
author="Miroslav {Skácel} and Igor {Szőke}",
title="BUT QUESST 2015 System Description",
booktitle="CEUR Workshop Proceedings",
year="2015",
journal="CEUR Workshop Proceedings",
volume="2015",
number="1436",
pages="1--3",
publisher="CEUR-WS.org",
address="Wurzen",
issn="1613-0073",
url="http://www.fit.vutbr.cz/research/groups/speech/publi/2015/skacel_mediaeval2015_submission_72.pdf"
}