Crossing obstacles in a walkway: on the capability of wavelet-based detection strategies using wearable sensor data

Mobility and its quality has direct and significant effect on quality of life. Passing over obstacles is unavoidable and its safe execution is a measure of mobility for community dwellers particularly for elderly and Parkinson patients with higher risk of falling. Algorithms for monitoring mobility in high risk people, need automatic detection to examine frequency and quality of passing over obstacles. Very few attempts can be found in the literature who just focus on the healthy population who need complex algorithms. Furthermore, in real life situations, people encounter a range of obstacle heights that should be detectable in such algorithms. In this paper a wavelet-based algorithm is examined and its performance is evaluated in detection of tall and short obstacles for two groups of healthy and Parkinson participants. Accuracy of this method was 98.5% for the 19 healthy elderly participants, and 90.6% for the 12 Parkinson patients. The maximum error in detection of obstacle crossing time was 0.1 second for either feet and for both barrier heights.