Multi-user human activity recognition through adaptively distinguishing location-independent individuals WiFi signal characteristics

The dataset accompanying this research endeavor serves as a valuable resource for scholars and researchers engaged in the domain of human activity recognition using WiFi signals, with a specific focus on multi-people sensing. The dataset was meticulously curated through experimentation conducted with a Raspberry Pi 4B setup, employing two distinct devices (RPi1, RPi2). The data collection spanned diverse scenarios and activities, encapsulating the complexities inherent in multi-user environments.

The dataset comprises CSI data, extracted using MATLAB code from TCPDUMP files, and subsequently subjected to a rigorous preprocessing and cleaning pipeline. This meticulous approach ensures the integrity and quality of the dataset, providing a robust foundation for subsequent analyses and experimentation.

In terms of experimental design, the dataset encapsulates a myriad of scenarios, each denoted by specific nomenclature (e.g., fx_S_2_1, fx_S_3_2), elucidating the nature of activities performed by individuals in a given environment. The experimental setup includes considerations for room dimensions (9m x 13m x 30cm), device placement (3m distance, 1m height for transmitter, 1.5m for receivers), and various activities such as standing, walking, sitting, running, and falls.

The dataset's academic significance lies in its capacity to facilitate in-depth investigations into the challenges and potentials of multi-people sensing using WiFi signals, particularly through the lens of Channel State Information. Researchers can leverage this dataset to validate and extend existing methodologies, fostering advancements in the realm of WiFi-based human activity recognition.

本研究配套的数据集，为从事基于WiFi信号的人类活动识别领域、且聚焦多人感知方向的学者与研究人员提供了极具价值的科研资源。该数据集通过基于树莓派4B（Raspberry Pi 4B）搭建的实验环境精心采集，共使用两台独立设备（RPi1、RPi2）。数据采集覆盖了多样化的场景与活动类型，充分涵盖了多用户环境中固有的各类复杂性。
本数据集包含信道状态信息（Channel State Information, CSI）数据：研究人员通过MATLAB代码从TCPDUMP文件中提取原始数据，随后经过严格的预处理与清洗流程。这一精细化处理流程保障了数据集的完整性与质量，为后续分析与实验提供了坚实的基础。
在实验设计层面，数据集涵盖了海量场景，每个场景均通过特定命名规则标识（例如fx_S_2_1、fx_S_3_2），用以说明对应环境中受试者的活动类型。本次实验的设置包含：房间尺寸为9m×13m×30cm，设备部署要求为发射端与接收端间距3米，发射端高度1米、接收端高度1.5米，同时覆盖站立、行走、就坐、奔跑以及跌倒等多种活动类型。
本数据集的学术价值在于，它能够助力研究人员深入探索基于WiFi信号的多人感知技术所面临的挑战与发展潜力，尤其可依托信道状态信息展开相关研究。研究人员可借助该数据集验证并拓展现有研究方法，推动基于WiFi的人类活动识别领域的技术进步。