AICOS-HAR: A human activity recognition dataset spanning multiple devices and body positions

The AICOS-HAR dataset is a large and diverse dataset that captures the movement of healthy subjects recorded in a controlled environment. This resource comprises 5,231 acquisitions (101.4 hours) from 106 participants, collected using 87 smartphones and 17 wearable devices, including accelerometer, gyroscope, magnetometer and barometer sensors. The activities recorded are walking, standing, sitting, running, lying down, and using stairs, ramps, and lifts. Earlier versions of this dataset have already enabled advances in position-invariant HAR [1][2], pedestrian navigation [3][4], and active learning [5], underscoring its utility.