Pervasive IOT: auto calibration in light sensor using embedded activity recognition
In the era of technology, lighting has become responsive and intelligent. When lighting is combined with internet of things (IoT), it could uphold greater innovated functionality. The emerging of smart lighting that support automatically switch on and off with additional embedded sensors and remotel...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Year Project / Dissertation / Thesis |
| Published: |
2021
|
| Subjects: | |
| Online Access: | http://eprints.utar.edu.my/4267/1/18ACB04706_FYP.pdf http://eprints.utar.edu.my/4267/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In the era of technology, lighting has become responsive and intelligent. When lighting is combined with internet of things (IoT), it could uphold greater innovated functionality. The emerging of smart lighting that support automatically switch on and off with additional embedded sensors and remotely control of light intensity has bring a comfortable and convenient lifestyle. However, the effect of color temperature on humans are neglected when the advancement of smart lighting is focusing on bringing convenient to humans. An incorrect application of colour temperature will lead to loss of productivity, damage vision ability and affect the well-beings. To address the issue, an intelligent smart lighting system that could change the lighting condition to the optimum colour temperature based on variety of activities is essential. The contribution of this project is to maximize the productivity of human in doing daily task, improved health, and protect the eye vision ability while replacing the conventional way of tuning the color temperature to autonomous.
In this project, a smart lighting system that able to change the light to optimal color temperature using embedded activity recognition is proposed. The dataset that used in this project is Moment in Time Dataset. A CNN-LSTM model that combining VGG-16 as a feature extractor and LSTM as classifier is proposed to perform activity recognition including dining, playing guitar as well as studying. The experimental results show that the proposed model behave well and achieves 75.45 % of precision and 77.76% of recall.
|
|---|