Fall detection has played a pivotal role in elderly healthcare. Traditional approaches of computer vision and wearable devices can be used to detect falling events to prevent heavy injuries. However, these approaches may raise some privacy concerns and inconveniences. WiFi sensing and machine learning combinations have recently been used for WiFi-based fall detection. In this study, we propose a low-cost system that can simultaneously detect human falling events and predict the falling location. Firstly, we introduce UIT-ESP32, a publicly available dataset of a person falling at various locations. The dataset was collected using low-cost ESP32 devices. Then, different machine learning models, including traditional and deep learning techniques, were applied to evaluate the system performance. We also introduce signal preprocessing methods, such as noise reduction and data segmentation, to enhance data quality and model performance. As a result, the LeNet model demonstrated exceptional precision, exceeding 99% for fall detection and localization. The experimental results highlight the potential for real-world healthcare applications of Wi-Fi-based systems as efficient tools, establishing a strong foundation for future human activity monitoring and innovation advancements.