Placement-Dependent Accuracy of a Smartphone-Based Sensor Application Compared to an Accelerometer-Based System for Measuring Physical Activity in Healthy Adults: A Validation Study

Accurately monitoring physical activity, including stationary cycling on an exercise bike, is important in managing chronic diseases and rehabilitation after lower limb surgery. This study aimed to validate a new smartphone-based sensor application (the BeSAFE+) for activity recognition and step counting across five phone placements, using the SENS Motion® system as a reference standard, and observed activity time as ground truth. In a laboratory-based study, 20 participants performed walking, brisk walking, running, high- and low-intensity cycling, sitting, standing, and lying activities while carrying five smartphones placed in the front and back trouser pockets, a backpack, a running armband, and a fanny pack, and wearing the activity tracker. The front pocket placement had the most accurate classification during cycling activities (89–93%) versus SENS Motion® (96–98%). For other activities, the highest overall classification accuracy was achieved with the phone in the back pocket. Overall, the SENS Motion® activity tracker demonstrated higher classification accuracy than most smartphone placements across all activities, except for running. Nevertheless, several smartphone placements and Application Programming Interface (API) approaches achieved activity recognition and step count estimates that were not significantly different from the SENS Motion® activity tracker, indicating that smartphone-based activity recognition can be valid under specific conditions.