search

akaraspt / deepsleepnet

DeepSleepNet: a Model for Automatic Sleep Stage Scoring based on Raw Single-Channel EEG
Apache License 2.0
398 stars 153 forks source link

About the Fs #39

Open hldqiuye123 opened 3 years ago

hldqiuye123 commented 3 years ago

I noticed that your flow chart of paper , the size of convolution kernel is defined as Fs/2 , but the model.py 77 Line was written that the real convolution kernel is 50. The SAMPLING_RATE = 256 in sleep_stage.py. Is the actual sampling rate 100 or 256? I will be looking forward to your reply. :)

besaman commented 3 years ago

the paper used 2 datasets: sleep-EDF with sampling rate 100, and MASS with sampling rate 256 ... so it depends on which one you use.

WaleedNazih commented 2 years ago

Thanks for the nice work.

If I have a dataset with a sampling rate of 256, should I modify the following parts model.py?

# Convolution
# network = self._conv1d_layer(input_var=input_var, filter_size=128, n_filters=64, stride=16, wd=1e-3)
network = self._conv1d_layer(input_var=input_var, filter_size=50, n_filters=64, stride=6, wd=1e-3)

######### CNNs with large filter size at the first layer #########
# Convolution
# network = self._conv1d_layer(input_var=input_var, filter_size=1024, n_filters=64, stride=128)
network = self._conv1d_layer(input_var=input_var, filter_size=400, n_filters=64, stride=50)