Ultra-Low Power Context Recognition Fusing Sensor Data from an Energy-Neutral Smart Watch

摘要: Today sensors and wearable technologies are gaining popularity, with people increasingly surrounded by “smart” objects. Machine learning is used great success in devices several real-world applications. In this paper we address the challenges of context recognition on low energy self-sustainable devices. We present an efficient multi-sensor system based decision tree to classify 3 different indoor or outdoor contexts. An ultra-low power smart watch provided a micro-power camera, microphone, accelerometer, temperature has been real field tests. Experimental results demonstrate both high mean accuracy 81.5 % (up 89 peak) consumption (only 2.2 mJ for single classification) solution, possibility achieve combination body worn harvesters.