Embodiments that relate to energy efficient gesture input on a surface are disclosed. One disclosed embodiment provides a hand-worn device that may include a microphone configured to capture an audio input and generate an audio signal, an accelerometer configured to capture a motion input and generate an accelerometer signal, and a controller comprising a processor and memory. The controller may be configured to detect a wake-up motion input based on the accelerometer signal. The controller may wake from a low-power sleep mode in which the accelerometer is turned on and the microphone is turned off and enter a user interaction interpretation mode in which the microphone is turned on. Then, the controller may contemporaneously receive the audio signal and the accelerometer signal and decode strokes. Finally, the controller may detect a period of inactivity based on the audio signal and return to the low-power sleep mode.