Estimation Accuracy of Average Walking Speed by Acceleration Signals: Comparison Among Three Different Sensor Locations

The ubiquity of wearable sensors now enables us to measure a user’s walking speed outside of a laboratory or clinical setting, during activities of daily living. However, this technology is recent, and researchers have yet to determine the locations on the body that produce the most accurate data from these sensors. This study aims to compare the accuracy of average walking speed estimation measured using acceleration data from three body landmarks: wrist, pelvis, and ankle. Estimation models are derived from the gait data of 247 healthy adults using stepwise linear multiple regression analyses. The absolute value of the within-participant mean of errors between actual average walking speed and estimated average walking speed is computed and compered across landmarks. The ankle is the most accurate locations from which to estimate average walking speeds from acceleration signals, whereas the wrist was the least accurate locations. Walking speed is an important measure of health and function, especially in older people, and accurately estimating walking speeds in daily life may be helpful in predicting health outcomes in the elderly.