Analysis of activity detection data pre-processing

Nowadays collecting sensor data is not just a trend, is a necessity. The increasing occurrence of smart home devices, which are now a permanent part of our homes and life, makes this process of gathering, storing and using this data crucial. One of the most important aspects is data pre-processing, as the process and methods involved must be carefully chosen to obtain the best results. We aim to improve data pre-processing for PIR based activity sensors. Our chosen activity detection system is CASAS, a complete system for detecting and storing activity, where we have access to raw data. The pre-processing is done in MatLab and we tackle the option to divide the main activity into multiple sub-activities or even segment the activity data into suitable windows. After that, we look over the data-labeling and feature selection process. Also, we analyze the possibility of making possible predictions based on existing data.