Fig. 3

Summary of global detection and segmentation algorithm. a Global workflow of the algorithm to detect the activities and transitions between activities using a motion capture system based on IMU. b Activities were detected by identifying peaks in selected IMUs that corresponded to different activities. \( \overline{\alpha} \) and \( \overline{\omega} \) denotes the normalized acceleration and angular velocity of the IMU. Multiple IMUs were used to provide complementary information that yielded more robust and accurate detection of activities. Here standing up was detected using the trunk a _z and the time derivative of the thigh a _y (\( {\overset{.}{\alpha}}_y \) >0). Similarly, sitting down was detected using the same IMU with (\( {\overset{.}{\alpha}}_y \) <0). c Segmentation was achieved by identifying the minimum/maximum to the left/right of the activity peaks. Multiple IMUs could be used to detect the same transition. The average time marked by these IMUs were used to estimate the beginning/ending of each activity. Here the sacrum and trunk angular velocities (ω_y) were used to estimate the transition that involved turning