基于弱特征匹配的后向散射非绑定式智能感知

Existing sensing technologies are most device-based, requiring the target carrying or being embedded with specific sensors or tags. Such a solution has some drawbacks, such as the constrain to users or objects, high deployment overhead, and limited application scenarios, and hence dose not meet the need of smart sensing in mobile and IoT environments. We propose a batch of device-free smart sensing methods based on weak feature extraction and matching. This project is based on commodity passive RFID systems and meets two requirements, detecting user behaviors and monitoring the amount of crowd. We leverage the backscattered physical signal of tags, focus on the techniques for extracting the characteristic of signals that reflects the targets’ behaviors and amount, such as the motion detection, nearby state monitoring, crowd estimation and feature extraction, etc. We plan to realize the effective feature mining and matching schemes for accurately identifying the user behavior and estimating the crowd population, and ultimately providing the practical, efficient, and reliable device-free smart sensing systems that are fully compatible with existing products and standards.

现有的感知技术主要采用绑定式方式，需要检测目标携带或植入特定的传感器或标签。这种绑定式的感知方式需要对目标进行约束，部署代价高，应用场景受限，无法满足对移动互联网、物联网等环境中对智能感知的需求。本项目在现有商用被动式无源射频识别系统基础上，提出基于弱特征提取和匹配的非绑定智能感知的方法，面向用户行为识别和群体数量监测两类基本应用需求，从后向散射的物理层信号入手，重点研究在复杂室内环境下，标签物理信号中可反映目标行为和数量的信号特征提取技术，如用户行为感知、标签临近状态探测、群体数量特征提取与估计等，进而实现能够精确反映用户行为和群体数量的特征挖掘及匹配机制与技术，最终为RFID行业提供与已有产品和标准完全兼容的，实用，高效，可靠的非绑定智能化感知系统。

本项目针对现有的绑定式感知技术存在的需对目标进行约束，部署代价高，应用场景受限等问题，面向移动互联网、物联网等环境中对非绑定式用户身份认证、行为识别、群体数量等基本应用需求，在现有商用被动式无源射频识别和Wi-Fi等无线系统基础上，提出基于弱特征提取和匹配的非绑定智能感知的方法，面向从后向散射的物理层信号入手，重点研究在复杂室内环境下，物理信号中可反映目标行为和数量的信号特征提取技术，如用户行为感知、临近状态探测、物体内部状态特征提取与估计等，进而实现能够精确反映用户行为和目标状态的特征挖掘及匹配机制与技术，为RFID等行业提供与已有产品和标准完全兼容的，实用，高效，可靠的非绑定智能化感知系统。课题组发表高水平学术论文17篇，其中CCF A类期刊6篇，CCF A类会议7篇，CCF B类会议1篇；授权专利3项。