Investigating Language Variability on the Performance of Speaker Verification Systems

Vaheb, Amir; Choobbasti, Ali Janalizadeh; Najafabadi, S. H. E. Mortazavi; Safavi, Saeid

doi:10.1007/978-3-319-99579-3_73

Cited by 2 publications

(2 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In their study, Chojnacka and colleagues [28] conducted multilingual experiments with a multilingual training dataset, which is not the main practice we want to investigate. The language dependency of the i‐vector technique has already been investigated [29, 30], but newer deep learning techniques have outperformed the older i‐vector technique. Fabien and Motlicek [31] investigated the performance of x‐vector models in forensic scenarios, but with acted speech, and the study did not focus on the effect of multilingualism (although the dataset was multilingual).…”

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Effects of language mismatch in automatic forensic voice comparison using deep learning embeddings

Schiferl

Fejes

2023

Journal of Forensic Sciences

View full text Add to dashboard Cite

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 Cllrmin 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.

show abstract

Section: Introductionmentioning

confidence: 99%