中文标题#
通过设备端少样本学习在可穿戴人体活动识别中连接泛化与个性化
英文标题#
Bridging Generalization and Personalization in Wearable Human Activity Recognition via On-Device Few-Shot Learning
中文摘要#
人体活动识别（HAR）使用可穿戴设备需要在不同用户之间具有强大的泛化能力，并且对个体进行高效的个性化处理。 然而，传统的 HAR 模型在面对用户特定的变化时往往无法泛化，导致性能下降。 为了解决这一挑战，我们提出了一种新的设备端少样本学习框架，该框架在可穿戴 HAR 中连接了泛化和个性化。 我们的方法首先在用户之间训练一个可泛化的表示，然后仅用少量标记样本快速适应新用户，在资源受限的设备上直接更新轻量级分类器层。 这种方法实现了计算和内存成本最小的稳健设备端学习，使其适用于实际部署。 我们在节能的 RISC-V GAP9 微控制器上实现了我们的框架，并在三个基准数据集（RecGym、QVAR-Gesture、Ultrasound-Gesture）上进行了评估。 在这些场景中，部署后的适应性分别提高了准确率 3.73%、17.38% 和 3.70%。 这些结果表明，通过无缝结合泛化和个性化，少样本设备端学习能够实现可扩展、用户感知和节能的可穿戴人体活动识别 \footnote {https://github.com/kangpx/onlineTiny2023}
英文摘要#
Human Activity Recognition (HAR) with wearable devices requires both strong generalization across diverse users and efficient personalization for individuals. However, conventional HAR models often fail to generalize when faced with user-specific variations, leading to degraded performance. To address this challenge, we propose a novel on-device few-shot learning framework that bridges generalization and personalization in wearable HAR. Our method first trains a generalizable representation across users and then rapidly adapts to new users with only a few labeled samples, updating lightweight classifier layers directly on resource-constrained devices. This approach achieves robust on-device learning with minimal computation and memory cost, making it practical for real-world deployment. We implement our framework on the energy-efficient RISC-V GAP9 microcontroller and evaluate it on three benchmark datasets (RecGym, QVAR-Gesture, Ultrasound-Gesture). Across these scenarios, post-deployment adaptation improves accuracy by 3.73%, 17.38%, and 3.70%, respectively. These results demonstrate that few-shot on-device learning enables scalable, user-aware, and energy-efficient wearable human activity recognition by seamlessly uniting generalization and personalization \footnote{https://github.com/kangpx/onlineTiny2023}.
文章页面#
PDF 获取#
抖音扫码查看更多精彩内容