📄 中文摘要
基础模型最近已扩展到时间序列领域，包括生理信号。然而，电皮肤活动（EDA）建模的进展受到缺乏大规模、经过策划且公开可获取数据集的限制。EDA 反映了交感神经系统的活动，广泛用于推断认知负荷、压力和参与度。目前，只有少数可穿戴设备提供连续、无干扰的感测，迄今为止唯一的大规模档案为专有数据。为了解决这一问题，研究团队编制了 EDAMAME，这是一个来自 24 个公共数据集的 EDA 轨迹集合，涵盖了来自 634 名用户的超过 25,000 小时的数据。利用这一资源，训练了 UME，这是首个专门针对 EDA 的基础模型。
📄 English Summary
A foundation model for electrodermal activity data
The recent expansion of foundation models into time series domains, including physiological signals, has faced challenges in electrodermal activity (EDA) modeling due to the lack of large-scale, curated, and openly accessible datasets. EDA, which reflects sympathetic nervous system activity, is commonly used to infer cognitive load, stress, and engagement. However, few wearable devices offer continuous and unobtrusive sensing, and the only large-scale archive available is proprietary. To address this gap, a collection named EDAMAME has been compiled, consisting of EDA traces from 24 public datasets, totaling over 25,000 hours from 634 users. This resource has enabled the training of UME, the first dedicated foundation model for EDA.