Author : Shafalii Sharma 1
Date of Publication :28th April 2023
Abstract: A cutting-edge method of detecting human behavior and sensing the environment is the use of Wi-Fi Channel State Information. By (re)using Wi-Fi routers, these techniques can be used for a variety of safety and security applications without the additional expensive hardware needed for vision-based approaches, which are also known to be especially privacy-invading. This study presents a complete pipeline for a Wi-Fi CSI-based system for identifying human activity that evaluates and contrasts two deep learning approaches. We examine the impact of various hardware setups on WiFi CSI transmissions. In order to provide more accurate evaluations of the model's classification performance, we contribute a novel and more realistic method of data gathering that seamlessly integrates the recognition of human activity in everyday life. We examine the performance of InceptionTime and LSTM-based classification models for identifying human behavior.
Reference :