Instantaneous walking speed estimation for daily life activity monitoring based on wrist acceleration
Benedikt Fasel, Farzin Dadashi, Kamiar Aminian · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · 2015
Walking is the main activity of daily life, and there is a great interest to characterize this activity by instantaneous walking speed and travelled distance per day. An ideal way to monitor daily activity is using accelerometer on the wrist like a watch. However, for estimating the walking speed, accelerations cannot simply be integrated because of sensor drift and arm movement artifact. Moreover, counting the total number of steps and multiplying it by a subject dependent coefficient is not practical as it requires a subject specific calibration and does not provide an instantaneous estimate of walking speed. In this study we train a linear regression model to obtain a calibration free estimate of the instantaneous walking speed based on the movement of the wrist.