Create download_s01_data.py

Datasets/download_s01_data.py

@@ -0,0 +1,56 @@
+# the data is downloaded from the original source, ie, 
+# http://gigadb.org/dataset/view/id/100295 (scroll down to Files tab, then click on '(FTP-Site)')
+
+# we use wget library to download from the ftp server
+# can be download using pip: pip install wget
+
+import wget
+
+
+# we're downloading only subject_01 data for now, but the url can be altered to download any of the 52 subjects' data
+url = 'ftp://parrot.genomics.cn/gigadb/pub/10.5524/100001_101000/100295/mat_data/s01.mat'
+wget.download(url)
+
+import scipy.io
+subject = scipy.io.loadmat('/content/s01.mat')
+
+# how subject.mat files are structured is desccribed here: 
+# https://academic.oup.com/gigascience/article/6/7/gix034/3796323
+left = subject['eeg']['imagery_left'][0][0][:64]
+right = subject['eeg']['imagery_right'][0][0][:64]
+event_markers = subject['eeg']['imagery_event'][0][0][0]
+
+# print(left.shape, event_markers.shape) # (64, 358400) (358400,)
+
+import pandas as pd
+import numpy as np
+
+df_left = pd.DataFrame(left).T.rename(columns = lambda x: 'left'+str(x))
+df_right = pd.DataFrame(right).T.rename(columns = lambda x: 'right'+str(x))
+markers = pd.Series(event_markers)
+df = df_left.join(df_right)
+df['mark'] = markers
+
+ind = list(df[df['mark'] == 1].index)
+ind = ind + [358399]
+intervals = list(zip(ind[:-1], ind[1:]))
+
+dfset_left = []
+dfset_right = []
+for start, end in intervals[:-1]:
+    dfset_left.append(df.loc[start:end-1, 'left0':'left63'])
+    dfset_right.append(df.loc[start:end-1, 'right0':'right63'])
+
+bci_left = []
+for df in dfset_left:
+    bci_left.append(df.values.T)
+bci_left = np.array(bci_left)
+np.save('bci_imagery_left.npy', bci_left)
+
+bci_right = []
+for df in dfset_right:
+    bci_right.append(df.values.T)
+bci_right = np.array(bci_right)
+np.save('bci_imagery_right.npy', bci_right)
+
+# print(bci_left.shape, bci_right.shape) # (99, 64, 3584) (99, 64, 3584)