Initial Findings:

Available Platforms:
1. Windows/Linux/Mac
2. Android
3. WP7

Potential Data Sources:
1. Wi-Fi
2. Bluetooth
3. Accelerometer/Gyroscope
4. Camera
5. HomeOS
6. GSM
7. GPS
8. Light Sensor
9. Internet
10. FM Radio

Data/Sensor per platform

Sensor	Windows	Linux	Mac	Android	WP7
Wi-Fi	Yes	Yes	Yes	Yes	???
Bluetooth	X	X	???	Partial*	No
GSM	No	No	No	Yes	Yes
Accelerometer	???	???	???	Yes	Yes
Camera	Yes	Yes	Yes	Yes	Partial**
HomeOS	Yes	Yes	Yes	Yes	Yes
Light Sensor	???***	???***	???***	Yes	No***
FM Radio	-	-	-	-	Yes

* Android's low level bluetooth functions are only available if you use the NDK and a rooted phone
** Most laptops have ambient light sensors, though I'm not sure if we can access them easily.
*** You can't silently/secretly get camera data on WP7 (according to this page) as an app runs; you have to launch a “CameraCaptureTask” which requires a user to take the snapshot himself.
**** All WP7 devices have a light sensor, but I don't think it's available to developers

Data Sources Pros/Cons:

• WiFi
• Pros
• Everywhere, supported by pretty much every device
• Can enumerate all nearby APs and check signal strength to estimate proximity
• Cons
• Battery drain
• Accuracy remains in question
• Bluetooth
• Pros
• Limited range = less interference? (compared to WiFi)
• Low cost of deployment?
• Cons
• Lack of APIs for phones (particularly WP7)
• Can't enumerate beacons or get RSSI easily (on phones)
• Accelerometer/Gyroscope
• Pros
• can probably use it to know when to start checking out other sensors/data sources
• Cons
• noisy data
• unclear as to how we'll match patterns or adjust to different people
• can't tell us on its own where the phone is, only when we can trust sensors
• Camera
• Pros
• potentially the most powerful way to figure out where we are
• Cons
• light in a room usually varies with time of day
• useless in low light (perhaps ignore cameras if light sensor sees no light)
• relies on camera being able to see a consistent scene
• not clear if object/scene recognition libraries exist for phones
• HomeOS
• Pros
• Provides access to sensors off the device, and access to the states of other devices (lights on, shades open, etc.)
• Cons
• available data depends on how extensive the HomeOS network is and what devices are plugged into it
• requires network connectivity (probably not a big deal)
• GSM (cell frequencies)
• Cons
• Only on phones, not very accurate on its own
• GPS
• Pros
• (can be) pretty accurate on its own
• Cons
• Doesn't work well indoors
• Batter drain
• only available on phones
• Long acquisition time (it's slow)

Calibration:

• WiFi
• Take device to each room's center and take WiFi “snapshot”
• Learn which controls to present over time by checking sensors every time user navigates to app controls that are associated with a particular room
• Have beacon in every room, strongest beacon heard tells device what room it's in
• Bluetooth
• Pretty much the same as for WiFi
• Accelerometer
• We'd need to have a way to measure a person's gait: see what their accelerometer trace looks like while walking and going up/down stairs and stuff
• Camera
• Train on unique objects in the room
• Check room's RGB profile (useful if room has unique wallpaper or something?)
• Possibility: combine with compass data to help differentiate snapshots
• HomeOS
• GSM
• I don't know how to calibrate this, but I don't think it'll be that useful in the context of the home. The phones have a localization API that uses this as part of their means to figure out where you are.
• GPS
• I don't think we can calibrate this one really. The phones have a localization API that uses this.