A Deep Learning-Based Human Identification System With Wi-Fi CSI Data Augmentation
Hyunggeun Mo, Seungku Kim
Abstract
Human identification systems generally include face recognition, iris recognition, radio frequency identification tags, and fingerprint recognition systems. However, these systems pose problems such as privacy violations, loss concerns, lighting requirements, and additional installation costs. Several studies have been conducted on human identification systems using Wi-Fi signals to address these problems. However, there exist problems such as a low number of identified per-sons, low accuracy, and high cost of data collection. In this paper, we present a deep-learning-based human identification system via Wi-Fi channel state information. To reduce the cost of data collection and increase the accuracy of human identification, we propose a data preprocessing and data augmentation process. They achieve an accuracy improvement of approximately 7%. In addition, we implemented one machine learning model and three deep learning models and demonstrated that the CLSTM model is suitable for the application through performance evaluation. The proposed system can identify up to 8 subjects with an accuracy of about 92%.