Bar of Soap

Predictive, grasp-sensitive handheld device

Overview

The Bar of Soap is a grasp-sensitive handheld device, aware of how a person is holding it. The device incorporates machine learning techniques that deduce the user's intentions from gestural and touch sensor input. The Bar of Soap was developed to explore new ways of interacting with handheld devices and off-device media.

The Bar of Soap research is led by Brandon Taylor, in collaboration with Jeevan Kalanithi, Matt Adcock, Quinn Smithwick and Dan Smalley. Developed in the Object-Based Media group at the MIT Media Lab.

Press

• Bar of Soap: "Grasp Sensitive" Device Guesses What You Want It To Do Next (Wired Gadget Blog, 12/03/2007).
• Seven Innovations That Will Change Your Life (PC Magazine, July 2007 Issue)
• "Bar of Soap" Prototype Detects Intent from Your Touch... (Boing Boing Gadgets, 11/28/2007)

Technology

The Bar of Soap prototype consisted of touch-sensing wrapping around the entire device as well as pattern recognition built for a low power, low speed embedded CPU.

• AVR microcontroller
• Bluetooth Radio
• ITO Transparent Capacitive Sensing
• 3 axis accelerometer
• Bistable LCDs
• MATLAB
• Statistical/Machine Learning techniques including:
  • Principal Components Analysis
  • Fisher Linear Discriminants
  • K-nearest neighbors
  • Parzen Windows
  • Bayesian Classification
  • General Linear Discriminants