Efficient classification of human activity using PCA and deep learning LSTM with WiFi CSI

Currently, we are entering the wearable internet era. It is easy to find network access points (APs) in today’s world. APs can be useful for more than just connecting to the internet. The waveform of WiFi signal changes when there is a person or human action between the two APs. In previous studies, we described how changes in waveforms affect the channel state information (CSI) of a signal and how machine learning can use this information to recognize and predict human behavior. In this study, we present points that can be improved on the previous study and the solution to filter the information. Preprocessing is one method for increasing performance. We found that using principal component analysis (PCA) reduced the processing time by 50% when classifying some classes. General–purpose PCA is useful for classifying data by reducing its dimensions.