Latest dataset & Tensorflow code for IEEE Communication Magazine.
Title: A Survey on Behaviour Recognition Using WiFi Channel State Information

Work by Siamak Yousefi, Hirokazu Narui, Sankalp Dayal, Stefano Ermon, Shahrokh Valaee

Tensorflow >= 1.0
numpy
pandas
matplotlib
scikit-learn

0. Download dataset from here.
   -> Notice: Dataset size is ~17GB
   -> You need to make new folder named "Dataset" and put every csv files into it.

1. "git clone" this repository.

2. Run the cross_vali_data_convert_merge.py
   -> This script makes csv files(input features & label) of each activity in "input_files" folder.　　

3. Run the cross_vali_data_convert.py
   -> This script makes learning curve images & confusion matrix in a new folder.　　

We collect dataset using Linux 802.11n CSI Tool.

The files with "input_" prefix are WiFi Channel State Information data.
-> 1st column shows timestamp.
-> 2nd - 91st column shows (30 subcarrier * 3 antenna) amplitude.
-> 92nd - 181st column shows (30 subcarrier * 3 antenna) phase.

The files with "annotation_" prefix are annotation data.