AccSleepNet

A deep learning model for automatic sleep stage scoring based on sum vector magnitude of accelerometer

Code for the model in the paper derived from DeepSleepNet: a Model for Automatic Sleep Stage Scoring based on Raw Single-Channel EEG by Akara Supratak, Hao Dong, Chao Wu, Yike Guo publication in arXiv.

before start the program, please specify the dataset folder in train.py then run the script.

Environment

• Windows 10 64bit
• CUDA toolkit 9.0 and CuDNN v6
• Python 3.6
• [tensorflow-gpu (1.6+)]
• matplotlib
• scikit-learn
• scipy
• pandas
• may others

Dataset

The first dataset is using V.T. van Hees et.al Estimating sleep parameters using an accelerometer without sleep diary

Scoring sleep stages

Current scoring method use cross-validation fold.

Get a summary

The train code will show a summary of the performance of AccSleepNet. The performance metrics are overall accuracy, per-class F1-score, and macro F1-score.

ToDo

-Change the network to intake the accelerometer data

Licence

• For academic and non-commercial use only
• Apache License 2.0