Improving Always-On Gesture Recognition Power Efficiency for Android Devices Using Sensor Hubs
Johann-P. Wolff, Sebastian Stieber, Tobias Rankl, Rainer G. Dorsch, Jens Rudolf, Christian D. Haubelt · 2016
In mobile user interaction scenarios, gesture recognition applications often process inertial sensor data and run continuously to reduce latency. In this paper, we evaluate real-life efficiency gains of gesture detection on-loading onto microcontroller-based sensor subsystems (sensor hubs). To this end, we implement smartphone wakeup gestures in two ways: (1) running on the Application Processing Unit (APU) that activates and unlocks the screen of an Android smartphone and (2) running on a sensor hub as interaction composite sensors that can directly wake up the APU through an interrupt line. We show that the sensor hub architecture can massively improve battery life in smartphones over the APU-based app while still enabling these new continuous sensing applications.