无线网络环境下的行为感知理论与关键技术

With the development of wireless network technology, 3G/4G mobile communication services and WiFi hotspots are virtually integrated into people's daily life. These network environments drive the emergence of a new paradigm of activity sensing. By utilizing the signaling data from cell towers, we can sense the space-time behaviors and social relationship of massive crowds, and reveal the temporal and spatial distributions and processes of crowds’ behaviors. Moreover, as individual activities also have specific effects on the channel state information of wireless signals, individual behaviors can be monitored by observing the changing patterns of signals. Based on this, we firstly plan to conduct research on the representation of activity sensing network. By representing the heterogeneous network as the lower dimensional vector embedding, large-scale mobility behaviors can be calculated and inferred. Secondly, we will study the theory behind individual activity sensing. To overcome the uncertainty caused by environmental factors, we will accurately calculate the quantitative relationship between sensing signals and human activities from the perspective of signal transmission. Meanwhile, we plan to explore the patterns of crowd behaviors, and excavate the semantics behind the spatial behaviors, which aims to solve the problem of lacking semantics of activities. Finally, on the basis of activity identification, we will detect the anomalous behaviors and conduct experiments to do validation. In summary, this project will expand the discovery paradigm of traditional activity sensing, provide solid theoretical and technical support for wireless sensing, and further promote the rapid development of corresponding applications.

随着无线网络技术的发展，移动3G/4G通信服务、WiFi热点已融入人们的日常生活，行为感知的新模式在这些网络环境下应运而生。例如，通过从基站中提取信令数据，可以感知海量群体的时空行为和社交关系，进而揭示社会群体行为的时空分布及过程。个体行为活动对无线信号的信道状态也存在特定影响，利用人体对信号的影响，可以对个体行为活动进行监控。基于此，本项目首先研究行为感知网络的表示，对异构网络的低维向量嵌入表示，使得大规模移动行为可以被计算推理。其次，研究个体行为感知理论，从信号传播的角度准确计算感知信号与人体行为的定量关系，应对环境等其他因素导致的感知不确定性。同时，研究群体行为模式，挖掘空间行为活动背后的语义，解决行为感知语义不丰富问题。最后，在行为识别的基础上，发现异常行为并进行实验验证。这些研究成果将扩展传统行为感知的发现模式，为无线感知提供坚实的理论和技术支撑，促进相关应用的快速发展。

随着无线网络技术的发展，各类无线信号（如WiFi，4G/5G，LoRa等）已经融入人们的生活，改善通信服务质量的同时，催生了无线网络环境下的行为感知新模式。本项目依托实际场景部署和采集的无线信号数据，对个体和群体行为感知开展深入研究，完成了下列研究工作：(1)无线感知理论研究方面，针对无线信号非接触行为感知领域缺乏系统性的基础理论，提出基于菲涅尔区衍射的人体感知模型，解决了定量计算感知信号与人体活动关系的问题。（2）开发了一系列基于无线信号感知的关键技术，重点研究了细粒度感知技术、远距离感知技术、多目标感知技术以及群体行为时空感知技术，为感知应用走向实际奠定重要基础。（3）在无线网络环境下的行为感知应用方面，构建了多个感知原型系统，包括细粒度呼吸监测系统，手势识别系统，日常健身行为识别系统以及群体人流移动行为预测系统。本项目的研究拓展了感知应用的深度和广度，提供了基于无线网络行为感知的新见解和新方法，特别是，多项研究成果具有实际应用前景。