Estimation of Walking Direction with Vibration Sensor based on Piezoelectric Device

• Published in: 2020 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) pp. 1 - 6
• Main Authors: Akiyama, Shinya, Yoshida, Makoto, Moriyama, Yumiko, Suwa, Hirohiko, Yasumoto, Keiichi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.03.2020
• Subjects: Estimation; Estimation of walking direction; LDA (Linear Discriminant Analysis); Legged locomotion; Logistics; Machine learning; Machine learning algorithms; Sensor systems; Vibration sensor; Vibrations
• DOI: 10.1109/PerComWorkshops48775.2020.9156202

Recently, device-free indoor positioning and tracking of persons is attracting attention. Some approaches use vibration sensors installed on the floor to identify the position of the vibration source by measuring and analyzing vibration strength and vibration fingerprints. However, vibration sensors can give the output information only on the strength of the vibration which shows that the vibration source is near or far from the sensor. Since the moving direction cannot be measured, it is difficult to apply the vibration sensor to track persons walking around in indoor space. In this paper, we propose a method for estimating the walking direction of a person using two vibration sensors. In the proposed method, we capture floor vibration using a pair of piezoelectric vibration sensors installed at a certain distance, calculate frequency domain features and estimate walking direction by machine learning. We conducted an experiment in a smart home where we installed our vibration sensors on the floor, asked four participants to pass near the sensors 20 times in two directions and estimated the walking direction by the proposed method. As a result, the walking direction was estimated with a maximum accuracy of 90%.