Wearable sensors improve activity recognition with smarter data augmentation
Coverage-Aware Virtual IMU Augmentation for Low-Resource Human Activity Recognition
Artificial Intelligence
Summary
Smartwatches and fitness trackers use motion sensors to understand what activities people are doing, but teaching these devices requires lots of examples, which are hard to collect. The authors created a way to make better 'fake' sensor data that fills in gaps where real examples are missing, making the training smarter and more reliable. Their method selects the most useful synthetic examples instead of randomly adding data, which helps the activity recognition work better across different people and situations. Tests on standard benchmarks showed their approach improved the ability to recognize activities compared to other methods.
What this means in practice
- •For wearable device developers: Improve training of activity recognition models using smarter synthetic sensor data to handle diverse users and limited labeled datasets.
- •For health monitoring teams: Enhance reliability of continuous activity tracking by incorporating selectively generated virtual sensor data during model training.
Authors
Jiayuan Gao, Yingwei Zhang, Ziyao Tang, Yuejia Ma, Yuanzhe Chen, Shuchao Song, Boshi Tang
Abstract
IMU-based human activity recognition (HAR) enables continuous, privacy-friendly monitoring of daily activities using wearable sensors. However, building reliable HAR models that generalize across diverse users and real-world conditions requires large amounts of labeled IMU data, which are expensive and difficult to collect. Existing approaches mainly rely on augmentation or synthesis to expand available data, but indiscriminately adding virtual samples may provide little new coverage and introduce unreliable supervision. To overcome these challenges, we propose a novel coverage-aware virtual IMU augmentation framework that decides where to supplement real data, how to generate and select virtual candidates, and how strongly to weight them during training. Specifically, we select diversity and scarcity anchors in a learned sensor embedding space, convert anchor dynamics into prompts, and generate virtual IMU candidates for each anchor. We then rank candidates by a selection cost combining anchor proximity and label consistency, and incorporate the selected candidates into HAR training with reliability-based weights. Experiments on public HAR benchmarks show that our method consistently improves recognition performance over competitive baselines, and ablation studies confirm the effectiveness of the proposed framework design.