Smartphone Collected CSI Dataset

This repository is the channel state information (CSI) dataset for CRISLoc: Reconstructable CSI Fingerprinting for Indoor Smartphone Localization, which is collected by Nexus 5 based on Nexmon.

If this dataset hepls your research, please cite the paper in your publications. : )

@article{gao2020crisloc,
  title={CRISLoc: Reconstructable CSI fingerprinting for indoor smartphone localization},
  author={Gao, Zhihui and Gao, Yunfan and Wang, Sulei and Li, Dan and Xu, Yuedong},
  journal={IEEE Internet of Things Journal},
  volume={8},
  number={5},
  pages={3422--3437},
  year={2020},
  publisher={IEEE}
}

CSI Data

The CSI data are stored in either research_laboratory_APX.mat or academic_building_APX.mat, where X is the index of APs, ranging from $1$ to $9$ or $10$ for the two scenarios. Within each file, we have a double matrix of number of data x number of subcarriers as collected CSI amplitude. We only present number of subcarrier as 49 for the 49 effective subcarriers. Note that all the CSI data are pre-processed as mentioned in the paper, including the frame filtering and CSI calibration (by RSS, and it can be negative).

Position Label

The position label are stored in either research_laboratory_position.mat or academic_building_position.mat. Within each file, we have a double matrix of number of data x 2, referring to the position of X axis and Y axis in meters. This position label are one-to-one mapping to the CSI data.

In this dataset, we mixed all the training/testing/reference data together and there is no difference for the labels in the figure.

Research Laboratory

First, we set up a testbed in the center of a $6m\times 15m$ research laboratory on the desks. There are few multipath reflections and little disturbance. RSS and CSI are collected at 90 positions with $0.5m$ spacing. $9$ APs are placed inside the room.

Academic Building

Then, we conduct the experiment on the third floor of an academic building. The test area is much more complex with many obstacles around. People walk around when fingerprints are collected, thus bringing disturbance to the data. The area covers an office, a corridor, and a lobby, and it is divided into grids with the edge width of $1.2m$ on average at the height ranging from $0.8m$ to $1.5m$. $10$ APs in all are deployed: $5$ in the corridor and $5$ in the office.