Context Acquisition Based on Load Sensing

1
Context Acquisition Based on Load Sensing

• Albrecht, Martin, Kristof, Adrian, Hans-W
• Lancaster University

Presenter Shih-yen 2005/04/19
2
Introduction

• Context-aware computing aims to embody
  information about the situation in which people
  interact with computational process.
• Load sensing though has some properties that
  appear to be well suited for context acquisition
  in ubiquitous computing.
• Related work SmartFloor

3
Determine 2-D position of objects on surface

• Four industrial load cells each can detect forces
  of up to 500N.
• The conditioning units are in turn fed to a
  standard 16-bit Analog to Digital Converter (ADC)
  which links to the serial line of the PC.
• each load cell can be read 4 times a second. The
  resolution of load sensing in this setup is
  approximately 16g.
• our driver software only calculates the position
  of objects between 100g and 100Kg (objects less
  than 100g are ignored).

4
Algorithm
5
Algorithm (cont.)
Table 135cm 75cm
6
Proof-of-concept

• Every day object
• 500ml water bottle (530g)
• Two litre water jar(2kg)
• the centre of the object as its position.
• Accuracy 2
• x 4cm
• y 2cm

7
Real Environment

• Tablecloth ,a TV-set, a book, a 2 litre water
  bottle, a magazine and a cable reel, totalling an
  overall weight of about 34kg.
• cm level accuracy.
• irrespective of
• objects already
• on the surface.

8
Detect Interaction Event

• d threshold of weight A average value
• D indicator of change
• Putting an object on the surface
• As dlt Ae
• Ds lt e
• Dm gt De
• Removing an object from the surface
• As d gt Ae
• The same
• Knocking an object over
• (DmgtF and Dm gtDs and Dm gtDe).
• As-Aeltd

9
Proof-of-concept

• 70 events, 94 correct, 6 missed, no event were
  misclassified.
• E1,E4 place on E2 knock over E3 remove

10
Living Laboratory
11
Application

• Object reminder
• Tracking position and estimate activities

12
Conclusion

• The weight information is useful context.
• The position of objects on the surface can be
  determined with high accuracy and reliability.
• The reported experiments also highlight the
  applicability in everyday settings
• future work
• identification, positioning and event detection
• develop algorithms that account for complex
  interactions
• with multiple objects on surfaces
• object tracking across multiple surface