| Summary: | With the emergence of Internet of Things (IoT) applications in smart homes, Human Activity Recognition (HAR) has acquired pivotal importance in this field. Conventional approaches to HAR include installing sensors in homes, wearable devices and Computer Vision solutions, but these approaches are not feasible in home applications due to their respective inconveniences. Studies have shown that the fluctuations in Wi-Fi signals can help reveal the movements of occupants. Therefore, in this project, Wi-Fi signals are exploited for classifications of human activities. This project underwent from collection of Wi-Fi signals (CSI frames), constructing and training deep learning models from scratch for such classification purposes. The classifier trained in this project can distinguish among seven different human activities with an accuracy of over 92%.
|
|---|