IEEE 802.11AX CSI Dataset for Human Activity Recognition

Citation Author(s):
University of Padova, Italy
Intel Corporation, USA
Intel Corporation, USA
Northeastern University, USA
Submitted by:
Francesca Meneghello
Last updated:
Thu, 04/27/2023 - 15:32
Data Format:
Research Article Link:
0 ratings - Please login to submit your rating.


The dataset includes channel frequency response (CFR) data collected through an IEEE 802.11ax device for human activity recognition. This is the first dataset for Wi-Fi sensing with the IEEE 802.11ax standard which is the most updated Wi-Fi version available in commercial devices. The dataset has been collected within a single environment considering a single person as the purpose of the study was to evaluate the impact of communication parameters on the performance of sensing algorithms. Three activities have been considered -- walking and running around the room, and staying in place -- and the data in the empty space scenario has also been recorded and made available as part of the dataset. The evaluation has been performed by using the state-of-the-art SHARP algorithm and is presented in the article "Toward Integrated Sensing and Communications in IEEE 802.11bf Wi-Fi Networks" accepted for publication in IEEE Communications Magazine.

The CFR data has been collected over an 80 MHz bandwidth channel on the 5 GHz band. For each available OFDMA sub-channel, the complex-valued channel data is recorded. Overall CFR data is available for the 996 OFDMA data sub-channels out of the 1024 sub-channels available (the remaining ones are control sub-channels where the channel reading is not available.

If you find the project useful and you use this dataset, please cite our article:

    author = {Meneghello, Francesca and Chen, Cheng and Cordeiro, Carlos and Restuccia, Francesco}, 
    journal={IEEE Communications Magazine},
    title = {{Toward Integrated Sensing and Communications in IEEE 802.11bf Wi-Fi Networks}},
    year = {2023},



The dataset contains data grouped into four classes: empty space (E), person walking (W), person running (R), person staying in place (S). Four captures for each class are available, lasting about two minutes each.


We set up an IEEE 802.11ax network with two Asus RT-AX86U Wi-Fi access points (APs). The network has been deployed in a house corridor by placing the routers along the two long edges, spaced apart by 4 m. The devices exchanged Wi-Fi data over the IEEE 802.11ax channel number 157 using the OFDMA resource unit RU1-996, i.e., with a bandwidth of 80 MHz and 996 sub-channels. We generated traffic with the iperf3 network analyzer tool setting an inter-packet distance of around 7.5 ms. The data has been collected through the AX-CSI tool that provides the CFR estimated on the packets collected by the receiver.

The adaptation of the SHARP algorithm to the IEEE 802.11ax data is available at The same GitHub repository contains the code to analyze the impact of the bandwidth and the sampling period on the performance of the Wi-Fi sensing algorithm.


A previous version of the dataset, with a higher amount of channel data, collected with IEEE 802.11ac devices can be found at