The repo contains the code for the paper "EEG-based Drowsiness Estimation for Safety Driving using Deep Q-Learning"
Since the paper is still under revision currently, so the preliminary version is submitted to arXiv for ease of understanding of the code.
Because GitHub constrains the maximum of uploading data file, the data files themselves are hosted in Google Drive, which could be accessed via: https://drive.google.com/open?id=10IUpTVcok-4KMIj0CV2lfvb4kRHE9Fq-
Currently only data of one subject are posted there, more in the following.
Prerequisites of the code are as follows:
- TensorFlow == 1.3 or 1.4, https://www.tensorflow.org/
- TRFL, https://github.com/deepmind/trfl
- dnn-lib, https://github.com/mingyr/dnn-lib
The eeg-double-dqn corresponds to the single-subject case, while eeg-dueling-dqn corresponds to the cross-subject case.
Feel free to contact the author for any question regarding it.