Automation of Electronics Device Control Based on Human Activity with Weighted K-Nearest Neighbor Classification

Abstract

Objective: Human daily activities such as turn off/on the lights, the air conditioner, the music player, etc., require a lot of time. These activities can be electronically automated, using human activity recognition. This idea is based on the idea that rational agents seek goals to increase their wealth. Method: In this research, it has been automated to turn on/off the light when someone passes through the road or stairs, turn on/off the air conditioner when someone is exercising, turn on/off the music player when someone is ironing clothes. The mechanization is done by using accelerometer and gyroscope sensors, to detect hand activities in these actions. Result: The sensor output in the form of a signal will be classified using the weighted K-Nearest Neighbor algorithm, so the system can classify what movement is being done. The results of the classification will activate the electronic device according to the purpose for which it was designed. The classification accuracy obtained is 95%, so it can help reduce of daily routine activities. Conclusion: The light switch used is a Smart Switch which is assembled using ESP-01 and a Relay, to turn off/on the switch. The Music Player used is coupled with WeMo’s D1Mini and DF Player Mini, to activate and deactivate the Music Player. The Universal Remote is used to turn on/off the AC.