深度神经说话人嵌入的类人跨语言泛化及其声学基础
原标题：Human-like cross-language generalisation in deep neural speaker embeddings and its acoustic foundations
AI 摘要
该研究考察深度神经说话人嵌入能否跨语言泛化，以及这种泛化与人类感知和声学结构的关系。作者使用粤语—英语双语语音，在语内和跨语言条件下从18个先进模型提取相似度，并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比较。结果显示，嵌入表现出与人类行为大体平行的跨语言泛化，反映准确的身份区分和对语言切换的敏感性，且这种泛化依赖大体共享但具语言敏感权重的声学线索，为跨语言语音相似性处理中的人机对齐提供了汇聚证据。
正文节选
Xu Li Jiang Dellwo Human-like cross-language generalisation in deep neural speaker embeddings and its acoustic foundations 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. Using Cantonese–English bilingual speech presented in within- and cro