A CSI-Based Human Activity Recognition Using Deep Learning.

Area of Science:

• Computer Science
• Healthcare Technology
• Signal Processing

Background:

• Human Activity Recognition (HAR) is crucial for healthcare, but existing methods like vision-based and sensor-based systems raise privacy concerns.
• The Internet of Things (IoT) and advancements in Information and Communications technologies have spurred interest in non-intrusive monitoring solutions.
• WiFi's ubiquity makes it a promising technology for intelligent daily activity monitoring, especially for elderly persons in healthcare settings.

Purpose of the Study:

• To develop and evaluate a novel Human Activity Recognition (HAR) system using WiFi Channel State Information (CSI).
• To demonstrate the effectiveness of CSI-based HAR as a privacy-preserving alternative to traditional methods.
• To assess the performance of a 2D Convolutional Neural Network (CNN) classifier for activity recognition using CSI data.

Main Methods:

• Collected CSI data for seven distinct human daily activities using a Raspberry Pi 4.
• Transformed CSI data into image representations for input into a machine learning model.
• Employed a 2D Convolutional Neural Network (CNN) for classifying the human activities based on CSI images.

Main Results:

• The proposed CSI-based HAR system achieved an accuracy of approximately 95% in recognizing seven different human activities.
• Experimental results indicated that the 2D CNN approach outperformed other methods such as 1D-CNN, Long Short-Term Memory (LSTM), and Bi-directional LSTM.
• The study successfully demonstrated the feasibility of using WiFi CSI for accurate and non-intrusive activity monitoring.

Conclusions:

• WiFi CSI-based HAR presents a highly accurate and privacy-conscious solution for intelligent activity monitoring in healthcare applications.
• The 2D CNN model effectively processes CSI-derived images for robust human activity recognition.
• This approach offers a significant advancement in leveraging ubiquitous WiFi infrastructure for elderly care and other HAR applications.