Datasets
Standard Dataset
D-Sense: Generalizing Datasets for Cross-Domain Wireless Sensing
- Citation Author(s):
-
ZhelunWang
Liaoning Technical University - Submitted by:
- Zhelun Wang
- Last updated:
- Thu, 11/28/2024 - 02:15
- DOI:
- 10.21227/zrmf-b356
- Data Format:
- License:
95 Views- Categories:
- Keywords:
Abstract
Generalized cross-domain sensing is a crucial step in driving the Internet of Everything. This dataset provides CSI information of Wi-Fi for different recognition tasks (gesture vs. gait) as well as DFS and (Absolute Distance Profile) ADP for researchers to validate the ADP. The ADP was tested on both the CNN-RNN networks that we utilized with parameter settings comparable to Widar 3.0, trained for 100 cycles. Then, we attach the confusion matrix for different tasks, which has been shown in the folder of the same level as the dataset, and you can refer to it for your reference. At the same time, we have attached a visualization example of ADP, which you can open and observe using MATLAB; the procedure is described in detail in the instructions.
We provide you with two datasets for two different sensing tasks, the Gait dataset (Gait) and the Gesture dataset (Gesture).
1. ADP-All. The folder includes the proposed characteristics of this paper; you can verify its reliability. The size of each file is 20*20*T tensor, where T is the time dimension (number of packets).
2. DFS-Gait/Gesture. This folder contains the DFS files that we extracted based on CSI.
3. CSI-Gait/Gesture. This folder contains the original CSI files used in the datasets we provide you. We use the CSI in the publicly available dataset for Widar 3.0 with configuration information consistent with Widar 3.0, and we provide you with descriptions of the datasets for the two sensing tasks, where the yellow labeled portion is the dataset we provide for you. (Related Datasets: widar3.0, doi: 10.21227/7znf-qp86)
4. ADP-Visual. The folder contains the visualization ADP example we prepared for you. It is the complete ADP (not scaled and downsampled). Its size is 120*120*T, where T is the time dimension (number of packets). You can use the imply in MATLAB for full-screen playback and viewing. Recommended imply parameters: Magnification: 800%; Color Map: parula(256); Frame Rate: 100 frames/sec.
Dataset Files
- Generalized Cross-Domain Perceptual Features: Absolute Distance Profiles (ADP) ADP-All.zip (1.08 GB)
- Gesture DFS Data DFS-Gesture.zip (23.51 GB)
- Gait DFS Data DFS-Gait.zip (12.63 GB)
- Gesture CSI Data CSI-Gesture.zip (4.72 GB)
- Gait CSI Data CSI-Gait.zip (2.51 GB)
- Visualization ADP ADP-Visual.zip (787.28 MB)
Documentation
| Attachment | Size |
|---|---|
 D-Sense Dataset Description | 142.83 KB |
| Description of the Gesture Dataset for Widar 3.0 | 394.33 KB |
| Description of the Gait Dataset for Widar 3.0 | 130.32 KB |
