Z-Tiles: Building Blocks for Modular, Pressure-Sensing Floorspaces Bruce Richardson, Krispin Leydon, Mikael Fernström, Joseph A. Paradiso
Introduction New Pressure- Sensitive Floorspace Successor to Litefoot (1998) & Magic Carpet (1997) Paradiso et al. (1997) The Magic Carpet: Physical Sensing for Immersive Environments. CHI97 Griffith and Fernstrom (1998) LiteFoot: A Floor Space for Recording Dance and Controlling Media. ICMC98.
Authors Bruce Richardson Interaction Design Centre, UL. Krispin Leydon Interaction Design Centre, UL. Mikael Fernström Interaction Design Centre, UL. Joe Paradiso MIT Media Lab, Boston
Requirement Scalable => Modular and interchangable
Requirement Managable => Self-organising and reconfigurable
Requirement Real-Time => Fast scanning and fast output Scan at 100 Hz Low latency data routing Minimal network overhead
Sensor Units
Z-Tile Circuitry
Z-Tile Architecture
Floor Prototype
Data Bottleneck Many wires vs. One wire Bottleneck at connection point
Options 1. Output only changed pressure readings 2. Group similar pressure readings
Options 1. Output only changed pressure readings 2. Group similar pressure readings
Blob Matching
Considerations 100 pressure scans/sec 10ms maximum computation time Minimum number of parameters Accurate blob matching Trade off Close fit vs Fewer parameters
Ellipse Matching Circles: Loose fit, few parameters Polygons: Tight fit, many parameters
Ellipses Ellipses - good compromise
Simple Matching
Average -> Centre
Axes -> Bounding Box
Resultant Ellipse
Results Pressures distilled to blobs 5 parameters per blob Implemented on a microcontroller >800 scans processed per second Rate achieved at 1/10 speed
Evaluation Processing time to spare Difficulty with diagonal blobs Therefore, look for better match
Angled Ellipses Pressure ReadingsWe have We want
General Method Determine centre as before Locate 2 most distant points Set this as major axis
General Method contd Rotate points so axis is horizontal Use bounding box to determine major and minor axes lengths
Results Implemented on microcontroller Integer-only calculations Lookups for rotations Computation time 1- 2ms
Summary Blob detection to reduce data 2 ellipse-matching algorithms Run time on hardware: <2ms
Future Work Integrate blob detection algorithm into tile software Detect blobs across tiles
Questions?