Why seated postures? - PowerPoint PPT Presentation

1 / 34
About This Presentation
Title:

Why seated postures?

Description:

Bilge Mutlu, Andreas Krause, Jodi Forlizzi, Carlos Guestrin, and Jessica Hodgins ... – PowerPoint PPT presentation

Number of Views:67
Avg rating:3.0/5.0
Slides: 35
Provided by: select8
Category:
Tags: bilge | postures | seated

less

Transcript and Presenter's Notes

Title: Why seated postures?


1
(No Transcript)
2
Why seated postures?
Automobile
Classroom
Wheelchair
Home
Office
3
Using posture information
4
Existing approaches
  • KinestheticMotion-capture markers or conductive-
    elastomer-embedded fabrics

Pellegrini and Iocchi., 2006
5
Existing approaches
  • KinestheticMotion-capture markers or conductive-
    elastomer-embedded fabrics
  • Vision-basedImage sequences from a single camera
    or multiple cameras

Tognetti et al., 2005
6
Existing approaches
  • KinestheticMotion-capture markers or conductive-
    elastomer-embedded fabrics
  • Vision-basedImage sequences from a single camera
    or multiple cameras
  • Pressure-sensing-basedPressure readings from the
    seating surfaces

Han et al., 2001
7
Challenges
  • Poor generalizationGood performance in
    classifying familiar subjects, poor performance
    with unfamiliar subjects due to high
    dimensionality.
  • High costHigh-fidelity pressure sensors are
    expensive.
  • Slow performanceProcessing high-fidelity sensor
    data demands computational power, which leads to
    slow processing.

8
Our solution
  • Robust generalizationUp to 87 accuracy in
    classifying 10 postures with new subjects.
  • Low costUsing 19 pressure sensors instead of
    4032. Reducing sensor cost from 3K to 100.
  • Near-real-time performance10Hz on a standard
    desktop computer
  • Novel methodologyUsing domain knowledge and
    near-optimal sensor placement.

9
Methodology
10
Learning Algorithm
  • Logistic RegressionSparse representation
  • Cross-validation10-fold, gender-balanced
    training and testing samples from different
    subjects
  • Separate setsTraining, testing, and reporting
    samples from 52 people in 5 trials
  • Implementation in Java

?
  • We would like to thank Hong Tan and Lynne
    Slivovsky for providing their data set for
    comparison.

11
Understanding pressure data
Modeling
12
Understanding pressure data
Modeling
13
Understanding our data
Modeling
14
Domain knowledge
Modeling
15
Features
Modeling
16
Features
Modeling
Classification accuracy
17
Separability test
Modeling
18
Feature elimination
Modeling
19
Methodology
20
Sensor granularity
Dimensionality Reduction
21
Sensor granularity
Dimensionality Reduction
22
How to place sensors?
Dimensionality Reduction
F
  • F, feature variables
  • V, locations and granularities
  • A subset A of V that maximizes information gain
    about F where H is entropy
  • NP-Hard optimization problem
  • We use near-optimal approximation algorithm

V
A ? V
IG(AF) H(F) - H(F A)
23
Near-optimal placement
Dimensionality Reduction
24
Sensor placements
Dimensionality Reduction
25
Near-optimal placement
Dimensionality Reduction
Classification accuracy
26
Methodology
27
Prototyping
28
Evaluation of prototype
  • 20 naive participants10-fold cross validation
    testing with 5 of the data
  • 78 accuracyIn classifying 10 postures
  • 10 Hz real-time performanceOn a standard desktop
    computer

29
Methodology
30
Conclusions
  • GeneralizabilityUp to 87 (with a base rate of
    10) achieved with unfamiliar subjects.
  • Low costHigher classification accuracy than
    existing systems using less than 1 of the
    sensors. 100 sensor cost compared to the
    commercial sensor for 3K (33 times reduction in
    price).
  • Near-real-time performanceAt 10Hz on a standard
    desktop computer.

31
Applications
Automobile
Classroom
Wheelchair
Home
Office
32
Future challenges
Next Steps
  • Transferring learning across chairsA
    transformation map could be created
  • Only static posturesTemporal dimension needs to
    be considered
  • The set of ten posturesThe set of postures
    should come from the activity

33
Summary of Contributions
  • A non-intrusive, robust, low-cost system that
    recognizes seated postures with generalizable,
    near-real-time performance.
  • A novel methodology that uses domain-knowledge
    and near-optimal sensor placement strategy for
    classification.

This work was supported by NSF grants
IIS-0121426, DGE- 0333420, CNS-0509383, Intel
Corporation and Ford Motor Company.
34
From Postures to Activities
Next Steps
  • Reading the paper
  • Watching TV
  • Reading paperwork
  • Watching TV eating
  • Sleeping
  • Talking on the phone
  • Reading a book
  • Craftwork
  • Reading the paper watching TV
  • Reading the paper eating
Write a Comment
User Comments (0)
About PowerShow.com