/LiteHAR

LiteHAR: LIGHTWEIGHT HUMAN ACTIVITY RECOGNITION FROM WIFI SIGNALS WITH RANDOM CONVOLUTION KERNELS

Primary LanguagePythonMIT LicenseMIT

LiteHAR

LiteHAR: Lightweight Human Activity Recognition from WiFi Signals with Random Convolution Kernels

Implementation of the LiteHAR model by Hojjat Salehinejad and Shahrokh Valaee.

The corresponding paper has been accepted for presentation at IEEE ICASSP 2022. Paper on ArXiv: https://arxiv.org/abs/2201.09310

Data

Here the link to the dataset used in the paper: https://github.com/ermongroup/Wifi_Activity_Recognition

Prerequisite

Python >= 3.6 numpy pandas scikit-learn numba joblib

How to Run

Run the bash script provided as: ./runner.sh

Parameters

Setup parameters in the runner.sh:

python3.6 main.py -m rigRocket -k 10000 -cv 1 -e 20 -i ../Dataset/Data/

where

  • i: path to the data
  • e: number of epochs (if necessary)
  • m: model
  • k: number of kernels
  • cv: number of cross-validation