Mobiquitous_2013_RFIDAR_LiangWang

Report
A Wearable RFID System for Real-time
Activity Recognition using Radio Patterns
Liang Wang1, Tao Gu2, Hongwei Xie1, Xianping Tao1, Jian Lu1,
and Yu Huang1
1
State Key Laboratory for Novel Software Technology, Nanjing University, P. R. China.
{wl,[email protected],{txp,lj,[email protected]
2
School of Computer Science and Information Technology, RMIT University, Australia.
[email protected]
Outline
• Introduction
• System Design
• Evaluation
• Conclusion & Future Work
State Key Laboratory for Novel Software
Technology, Nanjing University
2
Introduction - Applications
• Recognizing people’s activities continuously in real-time
enables a wide range of applications, e.g.,
Health Monitoring
Emergency Response
Entertainment
State Key Laboratory for Novel Software
Technology, Nanjing University
Assisted Living
3
Introduction - Motivation
• Traditionally, a body sensor network (BSN) is used to
capture activity data
Recognition
Algorithms
A BSN-based activity
recognition system
1.
2.
3.
State Key Laboratory for Novel Software
Technology, Nanjing University
Wearable sensors
Wireless communication
Processing unit
4
Introduction – Motivation
• Limitations of BSNs
• Human body affects the wireless link
quality
• Sensing, computing, storage,
communication devices
Packet loss
• Battery powered
State Key Laboratory for Novel Software
Technology, Nanjing University
5
Introduction – Related Work
• Passive RFID systems for localization [1] and gesture recognition [2]
• RSS patterns for localization and gesture recognition
• Advantages: cost-efficient, reliable, battery-free
• Limitations: fixed-reader & simple activities only
• Recent work on wearabe 2.4G network for human activity
recognition [3]
• Radio patterns for activity recognition
• Advantages: energy-efficient, amiable to packet loss
• Limitations: traditional BSN nodes
[1] S. Wagner, M. Handte, M. Zuniga, and P. J. Marron, “Enhancing the Performance of Indoor localization
Using Multiple Steady Tags,” Pervasive and Mobile Computing, vol. 9, no. 3, pp. 392–405, 2013.
[2] P. Asadzadeh, L. Kulik, and E. Tanin, “Gesture Recognition Using RFID Technology,” Personal and Ubiquitous
Computing, vol. 16, no. 3, pp. 225–234, 2012.
[3] X. Qi, G. Zhou, Y. Li, and G. Peng, “Radiosense: Exploiting Wireless Communication Patterns for Body Sensor
Network Activity Recognition,” in Proc. IEEE Real-Time Systems Symposium (RTSS), pp. 95–104, 2012.
State Key Laboratory for Novel Software
Technology, Nanjing University
6
Introduction – Our Approach
• Two observations
• There exists heavy attenuation of the human body to radio
communication band in which the UHF RFID operates
• RFID radio communication is highly affected by the tag-antenna
distance and orientation
• Intuition
Activities
Blockage of line-of-sight
Tag-antenna distance & orientation
Tag 1: RSS …
Radio Patterns
…
Recognition
Tag N: RSS …
Passive tag
UHF RFID reader
State Key Laboratory for Novel Software
Technology, Nanjing University
7
Introduction – Our Approach
• Research Issues
• How to discriminate different activities from the RFID
radio patterns?
• How to perform real-time activity recognition?
• Challenges
• False negative readings - a tag is in the antenna’s reading
range, but not detected; our current RFID reader can
activate one antenna at a time.
• Behavior difference - readings from different
combinations of tags and antennas may be different
even with the same condition.
State Key Laboratory for Novel Software
Technology, Nanjing University
8
System Design
• Antenna / Tag Placement
• 36 tags
• 9 body parts: both wrists, arms, legs, ankles, and the body
• 4 tags for each body part: reliable reading
• 4 antennas
• Detecting hand/arm movements: chest, back
• Detecting lower body movements: left feet, right feet
• Reading the tags
• 2 seconds for each antenna
• 8 seconds to complete a reading cycle
State Key Laboratory for Novel Software
Technology, Nanjing University
9
Preliminary Experiment
• Potential for activity recognition
C4.5
Recognition accuracy over 95%
State Key Laboratory for Novel Software
Technology, Nanjing University
10
System Design
• Data segmentation
• Fixed sliding-window of L seconds
• L is the application specific recognition delay bound
• Data completion – False negative readings
Temporal locality – tags recently detected are likely to be detected
again with similar RSS values
Last Window
Current Window
Current Data
Completed Data
Ant 0:
Ant 1:
Ant 2:
Combine
Data
Ant 3:
Time
State Key Laboratory for Novel Software
Technology, Nanjing University
11
System Design
• Feature extraction – Behavior difference
• Temporal features
• mean, variance, max, min, mean crossing rate, frequency domain energy, and
entropy of the RSS values for each pair of tag and antenna separately
• Spatial features
• the correlation coefficients of RSS series for different tags read by different
antennas
• Real-time recognition algorithm
• Online: recognition based on existing data
• Continuous: processing time < data collection time, i.e., L
• Solution: fixed sliding-window + SVM
State Key Laboratory for Novel Software
Technology, Nanjing University
12
Empirical Studies
• Data collection
• 4 volunteers - 8 activities - over 2 weeks
State Key Laboratory for Novel Software
Technology, Nanjing University
13
Empirical Studies
• Sliding-window size vs.
Recognition accuracy
• Real-time performance
93.6%
State Key Laboratory for Novel Software
Technology, Nanjing University
14
Empirical Studies
• Antenna and tag placement
State Key Laboratory for Novel Software
Technology, Nanjing University
15
Empirical Studies
• Transmission power level vs. Recognition accuracy
State Key Laboratory for Novel Software
Technology, Nanjing University
16
Conclusion
• We present in this paper
• Wearable UHF RFID-based recognition system
• Real-time recognition algorithm
• Future work
• Better sensing device
• Mobile phone integrated RFID reader
• More sensitive reader
• Better deployment strategy
• The minimal number of antennas and tags needed
• More empirical studies
• More activities
• More users
State Key Laboratory for Novel Software
Technology, Nanjing University
17
Thank you!
Q&A

similar documents