This code is the source code of our paper. Please refer to the paper for the specific parameters.
The dataset we used for model verification is from the PhysioNet/CinC arrhythmia detection challenge 2017, which contains 8,528 single-lead ECG recordings, each of which is derived from participants with length ranging from 9 seconds to 61 seconds. The data were sampled at 300 Hz (the shortest data has 2,714 data points and the longest data has 18,286).
Please using read_data.py to create the tfrecoed format data.
You can edit the file myscript.sh to adjust the hyper parameters of the training, and run entry.py to start running the program.
Set the is_test to True in myscrips.sh to test the model.
python==2.7 tensorflow==1.5.0 tensorlayer==1.8.3 CUDA==9.1 cudnn==7.3.1