* This repository was coded by myself and ateexD *
Presented at the Future of Information and Communication Conference (FICC) 2018, Singapore On April 6, 2018
-
This repository contains code to our research work/publication, Design Optimization Of Activity Recognition System on an Embedded Platform which primarily aims to design an Activity Recognition Engine that is optimized over computational complexity, power consumed and cost without compromising on the efficiency.
-
Optimization of computational complexity was carried out by reducing the number of features to be extracted. A set of seven simple time domain features were extracted to classify an activity.
-
The power consumption of the overall system was brought down by reducing the number of sensors used. The system was built with a single accelerometer in use which is used to get the data to compute and classify an activity. Thus bringing down the number of sensors to 1 also brings down the points of failure in the system.
-
The system was deployed on Raspberry Pi Zero ($5) which reduced the cost of the system. The selection of minimal number of sensors to contributes to this.