Device-Free Passive Identity Identification via WiFi Signals

Area of Science:

• Computer Science
• Electrical Engineering
• Signal Processing

Background:

• Device-free passive identity identification is crucial for applications like smart buildings and intrusion detection.
• Existing WiFi-based gait identification methods lack practical accuracy.
• Human gait possesses unique characteristics, similar to fingerprints, for individual recognition.

Purpose of the Study:

• To develop a novel device-free WiFi-based identity identification approach utilizing human gait.
• To enhance the accuracy of gait recognition using Channel State Information (CSI).
• To enable stranger recognition and precise identity identification in real-world scenarios.

Main Methods:

• Utilized Channel State Information (CSI) from WiFi signals for gait analysis.
• Applied Principle Component Analysis (PCA) and low-pass filtering to denoise CSI data.
• Extracted time and frequency domain gait features, selecting the most informative ones using information gain.
• Implemented stranger recognition with Gaussian Mixture Model (GMM) and identity identification with Support Vector Machine (SVM) using Radial Basis Function (RBF) kernel.

Main Results:

• Achieved effective stranger recognition and high identity identification accuracy.
• Demonstrated the system's performance on a dataset of over 1500 gait instances from eight subjects.
• Confirmed the system's low computational cost, making it suitable for practical deployment.

Conclusions:

• Wii offers a practical and accurate solution for device-free WiFi-based identity identification using human gait.
• The approach has significant potential for integration into home security systems.
• Further research can explore broader applications of CSI-based sensorless sensing.