A team at the University of Washington recently developed Allsee, a simple gesture recognition device composed of very few components. Contrary to conventional Doppler modules (like this one) that emit their own RF signal, Allsee uses already existing wireless signals (TV and RFID transmissions) to extract any movement that may occur in front of it.
Allsee’s receiver circuit uses a simple envelope detector to extract the amplitude information to feed it to a microcontroller Analog to Digital Converter (ADC). Each gesture will therefore produce a semi-unique footprint (see picture above). The footprint can be analyzed to launch a dedicated action on your computer/cellphone. The PDF article claims that the team achieved a 97% classification accuracy over a set of eight gestures.
Obviously the main advantage of this system is its low power consumption. A nice demonstration video is embedded after the break, and we’d like to think [Korbi] for tipping us about this story.