Effects of language mismatch in automatic forensic voice comparison using deep learning embeddings
Journal of Forensic Sciences · 2023-05 · 68 (3) · 871-883
DOI: 10.1111/1556-4029.15250 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Forensic Sciences, online erschienen am 31.03.2023. Verfasst von Dávid Sztahó und 1 weiteren Autor:innen. Beteiligte Einrichtung: Faculty of Electrical Engineering and Informatics, Department of Telecommunication and Media Informatics, Budapest University of Technology and Econo…. Hauptthemen laut MEDLINE: Deep Learning. Die Arbeit wurde bisher 19-mal zitiert (Crossref) und nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Nemzeti Kutatási Fejlesztési és Innovációs Hivatal.
Abstract (Originalsprache)
In forensic voice comparison, deep learning has become widely popular recently. It is mainly used to learn speaker representations, called embeddings or embedding vectors. Speaker embeddings are often trained using corpora mostly containing widely spoken languages. Thus, language dependency is an important factor in automatic forensic voice comparison, especially when the target language is linguistically very different from that the model is trained on. In the case of a low‐resource language, developing a corpus for forensic purposes containing enough speakers to train deep learning models is costly. This study aims to investigate whether a model pre‐trained on multilingual (mostly English) corpus can be used on a target low‐resource language (here, Hungarian), not represented by the model. Often multiple samples are not available from the offender (unknown speaker). Samples are therefore compared pairwise with and without speaker enrollment for suspect (known) speakers. Two corpora are used that were developed especially for forensic purposes and a third that is meant for traditional speaker verification. Speaker embedding vectors are extracted by the x‐vector and ECAPA‐TDNN techniques. Speaker verification was evaluated in the likelihood‐ratio framework. A comparison is made between the language combinations (modeling, LR calibration, and evaluation). The results were evaluated by Cllr min and EER metrics. It was found that the model pre‐trained on a different language but on a corpus with a significant number of speakers can be used on samples with language mismatch. Sample duration and speaking style also seem to affect the performance.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Faculty of Electrical Engineering and Informatics, Department of Telecommunication and Media Informatics, Budapest University of Technology and Economics, Magyar tudósok körútja 2, Budapest, 1117, Hungary.
- Doctoral School of Law Enforcement, Hungarian National University of Public Service, H-1083 Budapest, 2 Ludovika tér, Budapest, H-1441, Hungary.
ORCID-Profile: Dávid Sztahó
Förderung
- Nemzeti Kutatási Fejlesztési és Innovációs Hivatal
Zitieren
APA
Sztahó, D., & Fejes, A. (2023). Effects of language mismatch in automatic forensic voice comparison using deep learning embeddings. Journal of Forensic Sciences, 68(3), 871-883. https://doi.org/10.1111/1556-4029.15250
Vancouver
Sztahó D, Fejes A. Effects of language mismatch in automatic forensic voice comparison using deep learning embeddings. Journal of Forensic Sciences. 2023;68(3):871-883. doi:10.1111/1556-4029.15250
BibTeX
@article{Sztah2023,
author = {Sztahó, Dávid and Fejes, Attila},
title = {Effects of language mismatch in automatic forensic voice comparison using deep learning embeddings},
journal = {Journal of Forensic Sciences},
year = {2023},
volume = {68},
number = {3},
pages = {871--883},
doi = {10.1111/1556-4029.15250}
}
EUVIMED