研究人员为跨语言说话人验证（SV）系统开发了一个新的评估集，重点关注伊比利亚语种。该设置允许在一致的说话人身份下分析跨语言SV，解决了标准协议中将语言不匹配与说话人之间差异混淆的局限性。该研究将此应用于基于HuBERT的SV系统，发现虽然说话人差异会影响性能下降，但语言不匹配是跨语言性能损失的主要驱动因素。 AI
影响 为更精确地理解跨语言说话人验证系统中语言依赖性提供了依据。
排序理由 该集群包含一篇学术论文，详细介绍了特定AI研究领域的新方法和发现。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2607.01161v1 Announce Type: cross Abstract: Cross-lingual speaker verification (SV) systems typically exhibit performance degradation when enrollment and test utterances are spoken in different languages. However, standard evaluation protocols confound language mismatch wit…
Cross-lingual speaker verification (SV) systems typically exhibit performance degradation when enrollment and test utterances are spoken in different languages. However, standard evaluation protocols confound language mismatch with inter-speaker variability, as evaluation is gene…