PTrack: Enhancing the Applicability of Pedestrian Tracking with Wearables Yonghang Jiang, Zhenjiang Li, Jianping Wang Department of Computer Science City University of Hong Kong
6 Insurance companies Customer assessment Sedentary: 21 hours Active: 3 hours
Indoor navigation 124 shopping centers in HK (wikipedia) Huge market!!!
Performance?
10,000+ steps in 1 hour
Limitations • A rm’s motions Inaccurate step counter • Joint motion Inaccurate stride length estimation
Challenges • Identify interference activities. • Extract pure body’s motion. • User-specific profile self-training.
Rigid object Synchronized Matched Two motions Concurrent Not synchronized Offsets
Stride Estimation Extract body bounce b Existing model
Profile Self-Training Arm & leg lengths Search parameters Statistic property
Collaborate with existing works
Overall Accuracy 98% 98% 93%
Robustness
Stride Performance Average error: 5cm
Real Case Study G.truth: 141.5m PTrack: 136.4m
Thanks!
Recommend
More recommend