Overall System Design Goals

The goal of the Wockets project is to create a hardare and software system that allows for long-term physical activity measurement on mobile phones. This could be months or years in which people use the system continuously every day. This is an ambitious goal that requires considering many factors simultaneously in the design. Below are our assumptions, hard constraints, and desirable properties.

We started off with these facts and assumptions:
  • Today, the "typical" mobile phone has...
    • A fast CPU
    • A built in, 3-axis accelerometer capable of sampling at 10Hz+
    • Bluetooth wireless capability or a lower-power protocol, such as Bluetooth low energy
  • Some phones today allow continuous background processing, and in the future more will.
  • More than one accelerometer is needed to automatically detect many types of activities
  • It is not reasonable to ask people to carry their phone in one particular way (e.g., on the hip) for very long periods of time.
  • It will be possible to create software that entices people to remember to wear the sensors [How?]
  • There are many new applications for phones (for health, entertainment, communication, time management, etc.) that could be created using real-time detection of physical activity.

We aimed to meet these goals:
  • Performance:
    • 3+ external motion sensors able to communicate with the phone [Why?]
    • Accelerometer range between +/- 2 G and +/- 4 G [discuss]
    • Real-time activity recognition computed on the phone in real-time
  • Usability
    • 24h "swap once a day in the morning" performance -- no recharging motion sensors during the day [Why?]
    • Rechargeable
    • Sensor hardware small enough so that it does not attract undo attention and can be worn under clothing
    • Robust and automatic wireless connections - no manual pairing required
    • Safe
  • Housing:
    • No wires
    • Robust and sturdy for lengthy deployments of months or more
    • Comfortable to wear during all waking hours
    • Possible to wear sensors on upper body and lower body for adequate data collection
    • Easy to clean
  • Availability:
    • Affordable (i.e., substantially lower cost than current wireless accelerometers)
    • Easy to manufacture in small (prototype) and large (eventually) quantities
    • All designs open and free

We also hope the sensors/system have these attributes
  • Simple (improves robustness, lowers cost, encourages use, etc.)
  • Transmits raw data at high (20Hz + sampling rate)
  • Water proof (for shower/bath)
  • Wearable underneath clothing -- completely hidden
  • Easy to charge (i.e., connector extremely easy to insert and remove)
  • Secure -- nearby people cannot "grab" activity data
  • Scalable -- sensors work when other people wearing them are encountered
  • Tools needed for microprocessor (re)programming free or affordable
  • Easy to place on body, even for elderly who may have less hand strength and dexterity
  • Uses components that are in wide use and less likely to become obsolete
  • Designed with a possible change in phone default wireless protocol in mind

Finally, we also want the system to have these software attributes
  • Software runs on a variety of mobile phone platforms
  • Possible to add new activities on the fly
  • Operates continuously, but normal phone operations not impacted
  • Possible to output summary values equivalent to those used by other monitors (e.g., "counts")
  • Possible to determine difference between sensors not being used/carried at all and very sedentary behavior (e.g., sleeping or watching television)