Colorful PDR: Colorizing PDR with Shopping Context in Walking
Kohei Kanagu (Ritsumeikan University), Kota Tsubouchi, and Nobuhiko Nishio (Ritsumeikan University)
データサイエンス (Data Science) その他の取り組み (Misc.)
- The technology for detection of indoor locationinformation has been attracting much attention in the recentpast, since such information can be utilized for navigation andmarketing research. The commonly used indoor location detec-tion methods are Wi-Fi positioning, Bluetooth Low Energy (BLE)beacons and Pedestrian Dead-reckoning (PDR). PDR can providelocation information relative to the original position by analyzingacceleration and angular velocity using sensors in smartphoneswithout needing any positioning equipment in the environment.Meanwhile, to know the user’s context in the indoor environment,PDR, RFID and surveillance cameras are investigated. AlthoughPDR results can provide the atomic physical behavior such asthe direction of walking (forward/backward or right/left), it isdifficult to detect the user’s shopping behaviors to determineif he/she is approaching, searching or just browsing in theshopping venues. Our research objective is to estimate the user’swalking context appropriate for shopping venues with the samesensor information as used for conventional PDR. We propose anew method for walking context recognition that uses machinelearning with a feature vector of accelerometers and gyroscopes.Based on the experiments conducted, we found that this modelwas able to recognize the intuitive walking context. We alsoestablished the utility of our proposal in the experiment at theactual shopping venue and verified that it can be applied toimprove PDR positioning precision by 8%.