Sound based gesture recognition application.
This hack is based on a 2012 paper written by Microsoft's Research division. The application uses the speaker system to emit a continuous sound wave at 18kHz and samples the microphone at 44kHz. From the time-domain series data obtained from the microwave, we can perform a series of Fourier transforms and windowing functions to obtain the frequency-series data which gives us frequency and the amplitude of the sound wave at that frequency. We can recognize gestures by using Doppler shifts. When a sound wave is obstructed by an object (in this case that object is our hand), the sound wave is altered and this obstruction can be observed in the data. Different types of motion present themselves in different ways and we can determine the direction and velocity of the motion. This allows us to recognize gestures and motion with a high degree of accuracy. This application is not limited to the device. All it needs is a speaker and a microphone.
For a more detailed breakdown of the technology, check out this link