The TinyRadarNN datasets are a collection or RADAR data of hand gestures, collected Integrated Systems Lab at ETH Zurich using the Acconeer XR111 sensor. [1]
This project is brought to you by Moritz Scherer, Michele Magno, Jonas Erb, Philipp Mayer, Manuel Eggimann, Luca Benini of Digital Circuits and Systems Group of ETH Zurich (https://iis.ee.ethz.ch/), with the support of the Project Based Learning Center (https://pbl.ee.ethz.ch/).
The datasets are divided into one dataset with 5 gesture classes and one dataset with 11 gesture classes.
The dataset parameters are shown in the table below.
| Parameters | 5G | 11G (SU) | 11G (MU) |
| Sweep frequency | 256 Hz | 160 Hz | 160 Hz |
| Sensors | 1 | 2 | 2 |
| Recording length | 3s | 3s | 3s |
| # of different people | 1 | 1 | 26 |
| Instances per session | 50 | 7 | 7 |
| Sessions per recording | 10 | 5 | 5 |
| Recordings | 1 | 20 | 26 |
| Instances per gesture | 500 | 710 | 910 |
| Instances per person | 2500 | 7700 | 35 |
| Total instances | 2500 | 7700 | 10010 |
| Sweep ranges | 10-30 cm | 7-30 cm | 7-30 cm |
| Sensor modules used | XR111 | XR112 | XR112 |
License
The dataset is distributed under the Creative Commons Attribution Non-Commercial 4.0 (CC-BY-NC) license. All code is distributed under the Apache-2.0 license.
If you find this dataset useful in your research, please cite it with the citation below. You can find the associated paper on arxiv.org:
Citation:
@misc{scherer2020tinyradarnn,
title={TinyRadarNN: Combining Spatial and Temporal Convolutional Neural Networks for Embedded Gesture Recognition with Short Range Radars},
author={Moritz Scherer and Michele Magno and Jonas Erb and Philipp Mayer and Manuel Eggimann and Luca Benini},
year={2020},
eprint={2006.16281},
archivePrefix={arXiv},
primaryClass={eess.SP}
Contact:
For questions, please contact Moritz Scherer: scheremo@iis.ee.ethz.ch