WIFI信号状态信息 CSI 深度学习之数据集

Building occupant activity sensing dataset based on WIFI CSI（WiSA）

所有的数据以及实验参数都上传到了figshare中并配备详细说明，供参考。

论文链接：WiSA: Privacy-enhanced WiFi-based activity intensity recognition in smart buildings using personalized federated learning - ScienceDirect

In this experiment, WiFi signal collection was conducted in a standard bedroom (3.8m x 2.4m) and living room (3.2m x 4.4m).

Two Ubuntu 14.04 computers with Intel 5300 network cards and 3 external antennas were used, modified for Channel State Information (CSI) capture.

One computer functioned as a transmitter with a single antenna, and the other as a receiver with all antennas active. A 3×1 antenna setup was employed.

The 802.11n CSI tool captured 30 subcarriers per antenna link at a 20MHz bandwidth in the 5GHz band, with a transmission frequency of 1000Hz.

Devices were positioned at 1.3 meters height and spaced 2.2 meters and 3.2 meters apart, depending on room layout. All systems operated as per predefined parameters

15 volunteers (9 females and 6 males, aged between 23 and 26) were recruited for the experiment.

9 distinct activities were selected: lying down, reading, writing, walking, cleaning, arm training, squats, running, and jumping. These 9 activities were further categorized into 3 intensity levels: light, moderate, and intensity.

Each volunteer chose from the nine activities based on their usual living space habits and performed them in both the living room and bedroom. Each experimental session lasted 20 minutes, during which volunteers engaged in their chosen activities undisturbed.

The above is the description of the data source, please refer to the readme.txt for detailed file description.