Human-like cross-language generalisation in deep neural speaker embeddings and its acoustic foundations

Computer Science > Computation and Language arXiv:2611.05384 (cs) [Submitted on 2 Aug 2026] Title: Human-like cross-language generalisation in deep neural speaker embeddings and its acoustic foundations Authors: Tianze Xu, Xiyang Li, Xiaoming Jiang, Volker Dellwo.

计算机科学 > 计算与语言 arXiv:2611.05384 (cs) [提交于 2026 年 8 月 2 日] 标题：深度神经说话人嵌入中类人的跨语言泛化及其声学基础 作者：Tianze Xu, Xiyang Li, Xiaoming Jiang, Volker Dellwo。

Abstract: Generalisability is a central question for deep neural networks, including the cross-language generalisation of speaker embeddings. This study examined whether deep neural speaker embeddings generalise across languages and how such generalisation relates to human perception and acoustic structure.

摘要：泛化能力是深度神经网络的一个核心问题，其中包括说话人嵌入的跨语言泛化。本研究探讨了深度神经说话人嵌入是否能在不同语言间进行泛化，以及这种泛化如何与人类感知和声学结构相关联。

Using Cantonese—English bilingual speech presented in within- and cross-language conditions, we derived similarity from 18 state-of-the-art models and compared it with human ratings and acoustic features using linear mixed models and representational similarity analysis.

通过使用在语言内和跨语言条件下呈现的粤语-英语双语语音，我们从 18 个最先进的模型中导出了相似度，并使用线性混合模型和表征相似性分析将其与人类评分及声学特征进行了比较。

Embeddings exhibited cross-language generalisation that broadly paralleled human behaviour, reflecting accurate identity discrimination and sensitivity to language switching. This generalisation relied on largely shared acoustic cues with language-sensitive weighting, mirroring human patterns.

嵌入表现出的跨语言泛化能力在很大程度上与人类行为相平行，反映了准确的身份辨别能力和对语言切换的敏感性。这种泛化依赖于很大程度上共享的声学线索以及对语言敏感的加权，这与人类的模式相呼应。

Together, these findings provide converging evidence for human—machine alignment in cross-language voice similarity processing.

总之，这些发现为人类与机器在跨语言语音相似度处理方面的一致性提供了趋同的证据。