Abstract:We introduce LRLspoof, a large-scale multilingual synthetic-speech corpus for cross-lingual spoof detection, comprising 2,732 hours of audio generated with 24 open-source TTS systems across 66 languages, including 45 low-resource languages under our operational definition. To evaluate robustness without requiring target-domain bonafide speech, we benchmark 11 publicly available countermeasures using threshold transfer: for each model we calibrate an EER operating point on pooled external benchmarks and apply the resulting threshold, reporting spoof rejection rate (SRR). Results show model-dependent cross-lingual disparity, with spoof rejection varying markedly across languages even under controlled conditions, highlighting language as an independent source of domain shift in spoof detection. The dataset is publicly available at \href{this https URL}{\textbf{\underline{\textit{HuggingFace}}}} and \href{this https URL}{\textbf{\underline{\textit{ModelScope}}}}
From: Kirill Borodin [view email]
[v1]
Mon, 2 Mar 2026 20:11:06 UTC (4,225 KB)
[v2]
Fri, 3 Apr 2026 19:41:58 UTC (272 KB)
[v3]
Tue, 21 Apr 2026 22:46:15 UTC (272 KB)
[v4]
Sat, 27 Jun 2026 13:26:12 UTC (272 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。