Gesture Detection Using Doppler Sonar

While some computing devices include specialized hardware for gesture detection that enables hands-free operation, many commodity devices such as laptops do not include such hardware. This disclosure describes techniques, implemented with user permission, to automatically detect gestures performed in proximity of commodity devices at no additional cost by employing the on-device speakers and microphone as a Doppler sonar (sound navigation and ranging). The on-device speaker(s) generate an ultrasonic ping signal inaudible to the human ear and the device microphone(s) capture reflections from the user’s body parts. The type of Doppler shift in the reflected signal can indicate the direction of motion of the body part. The signal is provided as input to a trained classifier which can map the detected signal to the type of gesture the user is making. The described techniques can identify gestures quickly, thus enabling a smooth user experience for gesture-based interaction