StrengthSense: A Dataset of IMU Signals Capturing Everyday Activities with Diverse Physical Demands

Abstract

We present StrengthSense, a public full-body IMU dataset capturing everyday activities with diverse physical demands. StrengthSense contains approximately 8.5 hours of recordings from 28 subjects performing 13 activities, captured with 10 symmetrically placed 9-axis IMUs. The activities are grouped into four task-based categories representing distinct physical-demand profiles: Resistance, Load-bearing, Transition, and Baseline. This paper introduces the motivation, data collection protocol, dataset structure, and activity taxonomy, and provides representative analyses illustrating how the dataset can support wearable human activity recognition research. StrengthSense offers a benchmark for studying cross-subject recognition, sensor placement, reduced-sensor deployment, sensing-modality contribution, and physical-function monitoring. Dataset and code are available in https://doi.org/10.5281/zenodo.17424677 and https://github.com/yzy1997/multi-model-data-collector.