CSI Dataset for Wireless Human Sensing on 80 MHz Wi-Fi Channels

The dataset provides data for developing wireless sensing applications -- namely activity recognition, people identification and people counting -- leveraging Wi-Fi devices. The data is collected when the person/people move with the propagation environment. The human movements cause modifications to the multi-path propagation of Wi-Fi signals that reflect on the channel frequency response and, in turn, are identifiable by analyzing the channel readings.The dataset consists of Wi-Fi channel readings for up to seven activities and ten people. The dataset is collected via a monitor router implementing the Nexmon CSI tool that estimates the channel frequency response of ongoing Wi-Fi transmissions. The monitored traffic is generated by an IEEE 802.11ac-enabled Wi-Fi router that transmits data to a receiving device using an 80 MHz bandwidth channel. The data is collected in seven different environments, i.e., a bedroom, a living room, a kitchen, a university laboratory, a university office, a semi-anechoic chamber and a meeting room. Different respective positions and different hardware for the transmitter, receiver and monitor are considered to increase the domain diversity in the dataset. The dataset aims to provide a common ground truth for the development and comparison of wireless sensing solutions based on Wi-Fi. Overall, the dataset contains more than 13 hours of CSI recordings (for a total of 23.6 GB) among which 6 hours are for activity recognition, and 3 hours and a half are for people identification and people counting each.

本数据集面向无线感知应用开发提供数据支撑，具体涵盖活动识别、人员身份识别与人员计数三类任务，依托Wi-Fi设备完成数据采集。数据采集于人员在传播环境中移动的场景下，人体运动将改变Wi-Fi信号的多径传播特性，该变化会反映在信道频率响应中，进而可通过分析信道读取数据实现识别。本数据集包含最多7类活动与10名受试者的Wi-Fi信道读取数据。数据集通过搭载Nexmon CSI工具的监控路由器采集，该工具可对正在进行的Wi-Fi传输的信道频率响应进行估算。被监控的Wi-Fi传输流量由支持IEEE 802.11ac标准的Wi-Fi路由器生成，该路由器以80 MHz带宽信道向接收设备传输数据。数据采集于7种不同的传播环境，分别为卧室、客厅、厨房、大学实验室、大学办公室、半消声室以及会议室。为提升数据集的领域多样性，实验中采用了不同的设备摆放位置以及差异化的发射、接收与监控硬件配置。本数据集旨在为基于Wi-Fi的无线感知解决方案的开发与性能对比提供统一的基准真值。总体而言，本数据集包含超过13小时的信道状态信息（Channel State Information, CSI）录制数据，总容量达23.6 GB；其中6小时用于活动识别任务，人员身份识别与人员计数任务各占用3.5小时。