The Context Aware Cell Phone Project

Richard W. DeVaul, Steve Dunn.
MIThril, a borglab production. Richard W. DeVaul, Jonathan Gips, Michael Sung, Sandy Pentland
MIThril home

MIThril FAQ

MIThril wiki

Anduin window manager

MIThril Code

MIThril
hardware design

MIThril
project overview

The MIThril vision

MIThril diagrams

MIThril people

MIThril photos

Memory Glasses

MIThril Enchantment Software

MIThril
Real-Time Contxt Engine

MIThril IAP

Wearables Home Page

Cell phones represent the best and worst of modern wireless technology. Cell phones are one of the most popular tools of the early twenty-first century, facilitating business, recreation, and vital emergency communication. As the technology has become more popular, the social and safety problems associated with its wide-spread use have become more apparent. Cell phones are valuable tools, but they are also a significant distraction; The consequences of this distraction can range from socially inappropriate to life-threatening depending on the circumstances.

The Problem

We have all experienced the down-side of cell phone use; Cell phones ring in movie theaters, meetings, restaurants, and while we are in important face-to-face conversations. In some parts of the world restaurant and theater owners have gone so far as to employ cell phone jamming technology to limit the disruption caused by inappropriate cell-phone use. Cell phones are implicated in significant and increasing numbers automobile crashes each year, resulting in a national trend toward legislating restrictions on cell phone use while driving.

Partly this is a social problem; some people should learn to be more considerate in the use of this technology. But part of the problem is technological; A cell phone is deaf and blind, and depends completely on the user to manage its state. Providing users better tools to manage the state of their phones, tools which require less of their attention and can do more automatically, can only help. We can't stop inconsiderate people from using the technology in an inconsiderate way, but we can help responsible people make better and safer use of cell phones.

Context Awareness

If it were possible to build a phone that could determine the user's circumstances or context, this information could be used to change the phone's behavior in useful ways. Such a context aware phone could automatically switch profiles when the user enters a restaurant, sits in the driver's seat of a car, etc. In the case of phone management, nearly all of the complexity lies in the sensing and determination of context. Once the context is known, very simple rules can produce behavior that appears quite intelligent.

Context awareness requires sensing and inference to determine the user's context. Both the sensors and the inference techniques must be chosen carefully in order to make the best possible use of the limited real estate and computing power available in a portable form factor.

Sensors

We have chosen an array of sensors for the first context-aware cell phone prototype that allow us to determine important aspects of the user's state. All of these sensors are comparatively low-bandwidth except for the microphone, which we will band-limit to regions of interest for human speech. Managing bandwidth is important, because there is a direct correspondence between bandwidth, signal processing, and power consumption.
  • GPS receiver.
    The GPS receiver will provide outdoor location information
  • Precision three-axis accelerometer.
    The precision three-axis accelerometer provides information about the user's activity state (walking, standing still, etc.) and may additionally be used for gestural input.
  • IR tag reader and IR active tags.
    The combination of IR active tags and tag reader will be used to identifying important indoor locations (meeting room, office, etc) and to recognize special circumstances, such as sitting in the driver's seat of a car.
  • Microphone.
    The microphone will be used to recognize the user's voice and determine whether the user is in a conversation.

MIThril Inference Engine

The key to building a successful context-aware cell phone (or any other context aware application) is an effective context sensing and modeling system. The architecture of this system is quite important, since it must be powerful enough to do its job but simple enough to be implemented under tight resource constraints.

The result of our research into building a system with the required power and flexibility is the MIThril Inference Engine. The MIThril Inference Engine (MIE) provides a simple, flexible, and modular foundation for constructing context-aware applications. We are currently implementing the the MIE on our Linux-based MIThril computing cores, with the eventual goal of porting its core features to the Motorola i85s iDen phone.

Project Status:

November 15th

The context-aware cell phone project began in late summer 2001 with the donation of a Motorola i85s iDen phone and service by Motorola.

Since that time we have been focusing on sensor design, data acquisition, and the design of the MIThril Inference Engine that will enable us to implement the cell-phone management application. We have done preliminary work with the i85s hardware and Java interpreter, and verified that we are able to program the phone and talk to it over the serial port.

The first phase of the implementation will interface the i85s with the clothing-integrated MIThril wearable. The phone-management software will run as a MIE application on the Linux-based MIThril computing cores, and will communicate with a light-weight Java application on the i85s to change the phone's profile. This configuration will allow us to prototype the phone-management application with fewer bandwidth and computing constraints.

The second phase will integrate the i85s, sensors, and additional signal processing in a single package, resulting in a self-contained application.

We are working on the phase one implementation right now, and have a preliminary implementation of the motion, GPS, and tag-reader components working. Open problems include the Java interface for switching the phone profile and a light-weight speaker-identification model.