Human Detection Based on Learning and Classification of Radio Scattering Parameters and Para-Hermitian Eigenvalue Decomposition

Frank E. Ebong, Nicola Novello, Andrea M. Tonello

Official repository of the paper "Human Detection Based on Learning and Classification of Radio Scattering Parameters and Para-Hermitian Eigenvalue Decomposition" published at IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) 2024.

Algorithm for human detection based on a novel technique for pre-processing the scattering parameters and on a recently proposed new objective function for classification.

---

📷 Scenario

---

💻 How to run the code

Matlab part

Assume to have collected the S-parameters using a four-port Vector Network Analyzer (VNA). To pre-process the data using the Para Hermitian Eigenvalue Decomposition method, follow these instructions.

Python part

The directory where the scripts are must contain an additional folder Datasets containing 3 folders: Lambdas, Cauchy, and Raw. Lambdas and Cauchy contain the .mat files for the datasets of 0,1, and 2 people obtained using the corresponding pre-processing algorithms. Raw contains 3 folders (one for each class): Empty, Person, and Two_People that contain the s4p files obtained from the Matlab part.

The file main.py runs the experiments.

python3 main.py --mode Lambdas 

Where "mode" identifies the pre-processing algorithm used, which can be: Lambdas, Cauchy, No.

The files main_functions.py, classes.py, and utils.py comprise the needed methods and classes.

---

📝 References

If you use this code for your research, please cite our paper:

@inproceedings{ebong2024human,
  title={Human Detection Based on Learning and Classification of Radio Scattering Parameters and Para-Hermitian Eigenvalue Decomposition},
  author={Ebong, Frank E and Novello, Nicola and Tonello, Andrea M},
  booktitle={2024 IEEE 35th International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC)},
  pages={1--6},
  year={2024},
  organization={IEEE}
}

---

📋 Acknowledgments

The implementation is based on / inspired by:

• https://github.com/nicolaNovello/discriminative-classification-fDiv

---

📧 Contact

[email protected]

[email protected]