IndexMo: exploring finger-worn RFID motion tracking for activity recognition on tagged objects
- ,
- Shun Yao Yang,
- Bing Yu Chen
- National Taiwan University
Information
Output type
Original language
EnglishPages from-to (Number of pages)
Pages 129-134 (6 pages)Publication milestones
- Published - 09/09/2019
Publication status
Place of publication
New YorkPublisher
Association for Computing Machinery, Inc., United StatesISBN (Electronic)
978-1-4503-6870-4Host publication title
ISWC 2019 - Proceedings of the 2019 ACM International Symposium on Wearable ComputersRelated Event
Title
Description
Event type
ConferenceDate
09/09/2019 - 13/09/2019Location
Abstract
This work explores and evaluates the designs of finger-worn radio-frequency identification (RFID) motion tracking for activity recognition on tagged objects. We propose an index-finger-worn device that consists of a short-range (∼2cm) RFID reader and a pair of two inertial measurement units (IMUs), which are mounted at the locations where an artificial nail and a ring are worn. The short-range RFID reader recognizes the tagged object on finger touch, and then the IMU data are used for activity recognition. Data collected from the user of this device allows for a post-hoc analysis, which informs the activity recognition performance in various RFID+IMU and IMU-only configurations on the same task. The results of a ten-participant user study show that when the objects have similar physical form factors, the hybrid RFID motion tracking significantly outperforms the IMU-only tracking, especially in a larger-number set of objects. In our test, three IMU configurations (i.e., NailOnly, RingOnly, and Nail+Ring) achieved comparable action recognition performances, i.e., ≥90% accuracy, with 500 ms recognition time, though the NailOnly RFID+IMU configuration provided the highest wearability. The practical challenges toward a real-world deployment of a finger-worn RFID motion tracking system are also discussed.
