Rizk Hamada, Elmogy Ahmed
Computers & Control Engineering Department, Faculty of Engineering, Tanta University, Tanta 31527, Egypt.
Graduate School of Information Science and Technology, Osaka University, Osaka 565-0871, Japan.
Sensors (Basel). 2025 May 14;25(10):3108. doi: 10.3390/s25103108.
The deployment of autonomous AI agents in smart environments has accelerated the need for accurate and privacy-preserving human identification. Traditional vision-based solutions, while effective in capturing spatial and contextual information, often face challenges related to high deployment costs, privacy concerns, and susceptibility to environmental variations. To address these limitations, we propose , a novel AI-driven human identification system that leverages WiFi-based wireless sensing and contrastive learning techniques. utilizes self-supervised and semi-supervised learning to extract robust, identity-specific representations from Channel State Information (CSI) data, effectively distinguishing between individuals even in dynamic, multi-occupant settings. The system's temporal and contextual contrasting modules enhance its ability to model human motion and reduce multi-user interference, while class-aware contrastive learning minimizes the need for extensive labeled datasets. Extensive evaluations demonstrate that outperforms existing methods in terms of scalability, adaptability, and privacy preservation, making it highly suitable for AI agents in smart homes, healthcare facilities, security systems, and personalized services.
在智能环境中部署自主人工智能代理加速了对准确且保护隐私的人类身份识别的需求。传统的基于视觉的解决方案虽然在捕获空间和上下文信息方面有效，但往往面临与高部署成本、隐私问题以及易受环境变化影响相关的挑战。为了解决这些限制，我们提出了一种新颖的人工智能驱动的人类身份识别系统，该系统利用基于WiFi的无线传感和对比学习技术。该系统利用自监督和半监督学习从信道状态信息（CSI）数据中提取强大的、特定于身份的表示，即使在动态、多占用的环境中也能有效地区分个体。该系统的时间和上下文对比模块增强了其对人类运动进行建模的能力并减少了多用户干扰，而类别感知对比学习则最大限度地减少了对大量标记数据集的需求。广泛的评估表明，该系统在可扩展性、适应性和隐私保护方面优于现有方法，使其非常适合智能家居、医疗保健设施、安全系统和个性化服务中的人工智能代理。
