Skip to content

Human Activity Recognition based on WiFi Channel State Information

License

Notifications You must be signed in to change notification settings

Retsediv/WIFI_CSI_based_HAR

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

87 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Human Activity Recognition based on Wi-Fi CSI Data - A Deep Neural Network Approach

This is a repository with source code for the paper "Human Activity Recognition based on Wi-Fi CSI Data - A Deep Neural Network Approach" and respective thesis (it contains more details that are not covered in the paper).

Using Wi-Fi Channel State Information (CSI) is a novel way of sensing and human activity recognition (HAR). Such a system can be used in medical institutions for their patients monitoring without privacy violence, as it could be with a vision-based approach.

The main goal of this thesis was to explore current methods and systems which use Wi-Fi CSI, conduct experiments to analyze how different hardware configurations affect the data and possibility to detect human activity, collect the dataset and build the classification model for HAR task. 8 experiments were performed, the dataset in 3 different rooms was collected, and LSTM-based classification model was build and trained. We’ve shown the full pipeline of building Wi-Fi CSI based system.

Repository structure

  • router - contains source code for sendData and recvCSI. They are used to send data packet from one router and calculate the CSI data on another. Then recvCSI sends the data to a user computer via UDP connection for further processing.

  • data_retrieval - contains a program (run_visualization_server.py) that listens to recvCSI program, visualizes incoming data, and saves it to a file. Also, it has a script for a dummy server to emulate incoming data from the router (run_test_client.py) and a sample CSI data in binary format as it is coming from the router (data/sample_csi_packet_big_endian.dat).

  • model - has all the code for building the model and training it, scripts that were used to label activities, notebook for EDA, etc.

Dataset

The dataset can be downloaded by the following link.

Authors

  • Andrew Zhuravchak - Ukrainian Catholic University (UCU) former student
  • Oleh Kapshii - supervisor

License

This project is licensed under the GNU License - see the LICENSE.md file for details

Acknowledgments

Contributing

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.