Kalman-Filter-Based Walking Distance Estimation for a Smart-Watch

A novel walking distance estimation algorithm using the inertial sensors of the smart-watch is proposed. Firstly, the peaks of the norm of the accelerometer and gyroscope signals are detected. Due to arm swing, walking step detection using these peaks are not reliable. A Kalman filter is used to combine with the peak detection algorithm applied on the accelerometer and gyroscope norm peaks and robustly detect walking steps even if there is large arm swing. Walking distance is estimated using walking step time and walking length relationship. The proposed algorithm was tested on 25 subjects: each subject walked 50 m six times with different walking speed and different arm swing speed. The standard deviation of walking distance estimation error is 3.9 m (without person dependent calibration) and 1.9 m (with person dependent calibration) for a 50m distance.