A description of an accelerometer-based mobility monitoring technique

The two different posture thresholding methods were tested on long-term mobility data from one older adult subject. The subject did not have to follow a specific activity protocol during the recording period (4 days) and was shadowed by an observer in order to evaluate the accuracy of this technique. The monitoring hardware consisted of two accelerometer devices, one on the trunk and the other on the thigh and a pocket-sized ambulatory data-logger. Applying 'best estimate' thresholding, as opposed to mid-point thresholding, improved sitting detection accuracy by 18%, to 93% and lying detection accuracy by 5%, to 84%. Thus, based on these preliminary data, an accurate mobility monitoring system for older adults is described and it was observed that the actual posture threshold limits applied have a high impact on the mobility monitoring system's accuracy and are particularly important for accurately detecting postures when used over the long-term, in an uncontrolled environment.