SLIDE 1
WalkSense: Cla
lassify fying Home Occupancy States Usin ing Walk lkway Sensing
Elahe Soltanaghaei, Kamin Whitehouse
Department of computer science, University of Virginia
1 /18
Motion ≠ Occupancy
2 /18
Motion Occupancy 2 /18 Motion Occupancy 3 /18 Walkway Sensing 4 - - PowerPoint PPT Presentation
Motion Occupancy 2 /18 Motion Occupancy 3 /18 Walkway Sensing 4 - - PowerPoint PPT Presentation
WalkSense: Cla lassify fying Home Occupancy States Usin ing Walk lkway Sensing Elahe Soltanaghaei, Kamin Whitehouse Department of computer science, University of Virginia 1 /18 Motion Occupancy 2 /18 Motion Occupancy 3 /18 Walkway
SLIDE 2
SLIDE 3
Motion ≠ Occupancy
3 /18
SLIDE 4
Walkway Sensing
4 /18
SLIDE 5
Overview
Sleep Sleep
Bedroom Bedroom
Bathroom
Living room
Away Away Active Active
5 /18
SLIDE 6
Outline
- WalkSense: Occupancy detection algorithm
- Automatic training feature
- Evaluation
6 /18
SLIDE 7
WalkSense
Active-zone events
Sleep Walkway
Time
Candidate Transition Events
7 /18
SLIDE 8
Training WalkSense
By exploiting the WalkSense in offline mode How to train the models?
8 /18
SLIDE 9
Training WalkSense
- Candidate Sleep/Away Intervals: Inactivity longer than K
Sleep Walkway
Time
Candidate Sleep Intervals Candidate Away Intervals
< K
Active-zone events
Outside Walkway
Automatic Labeling No user involvement Incremental Training Improve detection model over time
9 /18
SLIDE 10
Experimental Setup
- 350 days worth of data
Active Sensing Zone Sleep Walkway Zone Outside Walkway Zone
Kitchen bedroom
House A
Living room
House C, D
Kitchen bedroom Living room bathroom Kitchen bedroom bedroom
House B
Living room
House E
Kitchen bedroom Study room Living room Kitchen bedroom
Aruba
Living room
10 /18
SLIDE 11
Experimental Setup
- Baseline: HMM-base
- Evaluation Metrics
- Energy penalty: away or sleep periods are wrongly detected as Active
- Comfort penalty: Active periods are wrongly detected as Sleep or Away
11 /18
SLIDE 12
Experimental Setup
61% of the day outside home, or sleep
12 /18
SLIDE 13
Evaluation
30% lower energy penalty 32% lower comfort penalty
13 /18
SLIDE 14
Evaluation
95% vs. 83% Accuracy
Actual Actual Predicted Predicted
WalkSense HMM Baseline
14 /18
SLIDE 15
Analysis
Training Size Effect
15 /18
SLIDE 16
Conclusion
- WalkSense: Occupancy Detection algorithm based on Walkway Sensing
- Higher Accuracy
- Fewer Sensors
- 96% Accuracy with 30% and 32% lower energy and comfort penalty
- Limitation: Assumes direct relationship between
Occupancy States & Occupancy zones.
- TV
- Study Desk
16 /18
SLIDE 17
Conclusion: Generalizable
17 /18
SLIDE 18
Thank You!
18