Datasets
Standard Dataset
A CSI Dataset for Wi-Fi-based Personnel Identity Recognition
- Citation Author(s):
- Submitted by:
- Jichen Bian
- Last updated:
- Mon, 05/13/2024 - 11:53
- DOI:
- 10.21227/7yrm-ys36
- License:
325 Views- Categories:
- Keywords:
Abstract
This dataset utilizes Asus RT-AC86U routers and nexmon tools to collect Channel State Information (CSI) data in a 7 by 5 meters meeting room furnished with typical furniture including a conference table, several chairs, and a locker. The data, stored in .pcap format, is accompanied by processing code on GitHub, enabling parsing into CSI matrix data stored in .npy format. Each CSI matrix contains amplitude and processed phase values for four channels, encompassing data from both external and internal antennas within the room. These matrices facilitate in-depth analysis of dynamic changes in the wireless channel induced by human movement, serving as a resource for research in human activity recognition and related fields.
Wireless communication has become ubiquitous in modern environments, supporting a wide range of applications from mobile devices to IoT devices. Understanding the behavior of the wireless channel is crucial for optimizing communication systems, particularly in scenarios where human activities introduce dynamic changes to the channel characteristics.
In indoor environments such as meeting rooms, where the users may move around, the wireless channel experiences variations due to multipath fading, shadowing, and interference. CSI provides valuable insights into these variations by capturing the amplitude and phase of the signals transmitted between devices.
This dataset presents a comprehensive collection of CSI measurements captured in a real-world meeting room scenario. By deploying Asus RT-AC86U routers equipped with nexmon tools, we obtained high-resolution CSI data capturing the dynamic changes in the wireless channel induced by human movement.
The meeting room setting, with its typical furniture arrangement and dimensions, reflects a common environment encountered in everyday scenarios. The routers' placement within the room and the periodic single-antenna data transmissions, along with continuous monitoring with four antennas, allowed us to capture a diverse range of channel conditions.
Each CSI measurement in the dataset represents a snapshot of the wireless channel, comprising amplitude and phase values across multiple subcarriers and discrete time slots. The dataset includes measurements from multiple participants moving randomly within the room, providing a rich source of data for studying the impact of human activity on the wireless channel.
Researchers can leverage this dataset to develop and evaluate algorithms for human activity recognition, device localization, and dynamic channel modeling. Additionally, the availability of processing code facilitates easy access to the raw data, enabling further analysis and experimentation in the field of wireless communication and signal processing.
Additionally, to facilitate effective utilization of this dataset by researchers, preprocessing code is provided, accessible via GitHub. This experiment was set up with an 80MHz bandwidth and a data transmission interval of 10ms. The preprocessing code aids researchers in extracting and converting raw .pcap files into .npy format CSI matrices for further analysis. After preprocessing, the default dimensions of a single CSI matrix are (4, 256, 256), where 4 represents the number of receiving antennas, the first 256 represents the number of subcarriers, and the second 256 represents the time slots. The preprocessing code can be downloaded from the following link: https://github.com/BBJJCC/Preprocessing-code.
Furthermore, information regarding the participants, including their gender, age, height, and weight, is available in the "information.xlsx" file. This additional data can be utilized for deeper analysis and understanding of how different physical characteristics might affect the CSI measurements.
