Dysun - An activity-aware smart IOT light using human activity recognition

• Main Author: Peh, Hong Bo
• Format: Final Year Project / Dissertation / Thesis
• Published: 2023
• Subjects: H Social Sciences (General); T Technology (General); TD Environmental technology. Sanitary engineering
• Online Access: http://eprints.utar.edu.my/6040/1/fyp__CS_2023_PHB.pdf; http://eprints.utar.edu.my/6040/

This is a development-based project to develop an automated smart lighting system that integrated a human activity recognition (HAR) model using computer vision. In recent years, smart homes have gradually become more popular in people's daily lives. An increasing number of smart home products have been introduced to the market and are widely embraced. Such market trend reflects people's demand for enhancing their quality of life by making use the advancement of technology in practical and real-life scenario. As one of the essential elements within a house, lighting systems have consistently been a popular cornerstone in the development of smart home systems to fulfil the requirement of flexible control and comfort use. However, none of the existing system could realise pure automation without any explicit involvement of human control. To fulfil the absence of such product, this project proposed an automated smart lighting prototype that apply human activity recognition model which work as the “brain” of the lighting system to understand what human is doing and to adjust the lighting condition accordingly. The prototype was built with a CSI camera, Jetson Nano and Yeelight smart light bulb to demonstrate the lighting system. A CNN-LSTM HAR model with evaluation accuracy at 74% is used as the backbone for activity recognition.