SensorLib: an Energy-efficient Sensor-collection Library for Wear OS

In recent years, wearable technology has gained popularity due to features like long battery life, network connectivity, and fitness monitoring. Human Activity Recognition has emerged as a popular use case for smartwatches, enabling the recognition of activities starting from internal sensors. Data acquisition from sensors is crucial in wearable devices because if not properly implemented can reduce battery life or device responsiveness. The paper presents an energy-efficient programming library for real-time sensor sampling on smartwatches using native Wear OS sensor APIs. The library's implementation is evaluated on a real smartwatch for code size, memory utilization, and power consumption. The preliminary results empirically demonstrate that the solution proved to be light and versatile enough to be used on wearable devices without heavily compromising battery life and system performance.