Sensors & activity recognition
A fundamental challenge for much of the “Connected Home and Community” activity on SiDE is reasoning about what people are doing at any moment in time. This is an open problem in computer science known as activity recognition. To do this in practice we need tiny, very sensitive, sensors that can be both worn and embedded in the world around us. SiDE is producing a suite of open source hardware and software for activity recognition that both underpins our own research and lowers barriers for other researchers around the world.
SiDE has developed two accelerometer-based sensors, AX3 and WAX3 (i.e. sensors that can measure acceleration). These are being used in a number of different ways. For example, we have embedded sensors in knives, and use activity recognition algorithms to detect how these knives are being used (e.g. chopping). Worn accelerometers allow us to estimate levels of energy expenditure, sleep and other behaviour.
AX3 logs data over long periods of time (up to 21 days) and stores the data on the device (good for longer term field studies); whereas WAX3 transmits data continuously over a wireless data connection (good for interactive applications). We are also developing MOVeCLOUD, an eScience platform that allows the massive quantities of data the sensors generate to be stored, processed and shared by researchers.
Cas Ladha (hardware)
Dan Jackson (software)
The sensor hardware and software team have recently developed a new platform for augmented tangible objects on FTIR tabletops; for more details see the poster TouchBridge: augmenting active tangibles for camera-based multi-touch surfaces presented at ACM Interactive Tabletops and Surfaces (ITS’10)
OpenMovement code is available at: http://code.google.com/p/openmovement/