def read_a_mat(eg_data, label):
'''
given clips eeg data of one person, preprocess and merge them
'''
# Final Data to return
train_data, train_label, test_data, test_label, \
validate_data, validate_label = [], [], [], [] , [], []
# Loop All Clips of A man
for i in range(15):
all_sample_number, all_samples = [], []
for key_word in ['de_LDS', 'de_movingAve']:
A_key = key_word + str(i + 1)
clip_a_data = eg_data[A_key]
_, b, _ = clip_a_data.shape
all_sample_number.append(b)
# Min Sample
min_sample = np.min(all_sample_number)
for key_word in ['de_LDS', 'de_movingAve']:
A_key = key_word + str(i + 1)
clip_a_data = eg_data[A_key]
clip_a_data = np.transpose(clip_a_data, (1, 0, 2))
clip_a_data = clip_a_data[:min_sample, :, :]
all_samples.append(clip_a_data)
# A Clip Data
clip_data = vstack_list(all_samples)
clip_label = label[i] + 1
clip_label = clip_label * np.ones([int(len(clip_data)), 1])
# print('\tClip %d shape:' % (i + 1), clip_data.shape)
# Train and Test
if (i + 1) in train_clips:
train_data.extend(clip_data)
train_label.extend(clip_label)
elif (i + 1) in test_clips:
test_data.extend(clip_data)
test_label.extend(clip_label)
elif (i + 1) in validate_clips:
validate_data.extend(clip_data)
validate_label.extend(clip_label)
# Join data
one_train_data = np.transpose(np.array(train_data), (0, 2, 1))
one_train_label = np.array(train_label).reshape(-1, 1)
one_test_data = np.transpose(np.array(test_data), (0, 2, 1))
one_test_label = np.array(test_label).reshape(-1, 1)
one_validate_data = np.transpose(np.array(validate_data), (0, 2, 1))
one_validate_label = np.array(validate_label).reshape(-1, 1)
# print(np.array(one_train_data).shape, np.array(one_test_data).shape, np.array(one_validate_data).shape)
# print(one_train_label.shape, one_test_label.shape, one_validate_data.shape)
return one_train_data, one_train_label, one_validate_data, \
one_validate_label, one_test_data, one_test_label
def load_all_person(label, tmp_file):
'''
Load all person's eeg files
Fetch all data for train, test, validate
'''
all_train_data, all_train_label, all_test_data, all_test_label, \
all_validate_data, all_validate_label = [], [], [], [], [], []
one_train_data, one_train_label, one_validate_data, one_validate_label, \
one_test_data, one_test_label = read_a_mat(tmp_file, label)
print('One person Shapes:', one_train_data.shape, one_validate_data.shape, \
one_test_data.shape)
all_train_data.append(one_train_data)
all_train_label.append(one_train_label)
all_validate_data.append(one_validate_data)
all_validate_label.append(one_validate_label)
all_test_data.append(one_test_data)
all_test_label.append(one_test_label)
all_train_data = vstack_list(all_train_data)
all_train_label = vstack_list(all_train_label)
all_validate_data = vstack_list(all_validate_data)
all_validate_label = vstack_list(all_validate_label)
all_test_data = vstack_list(all_test_data)
all_test_label = vstack_list(all_test_label)
# Shuffle
all_train_data, all_train_label = shuffle_train(all_train_data, all_train_label)
print('Final data shape', all_train_data.shape, all_test_data.shape, all_validate_data.shape)
return all_train_data, all_train_label, all_validate_data, \
all_validate_label, all_test_data, all_test_label
def load_all_person(label):
'''
Load all person's eeg files
Fetch all data for train, test, validate
'''
all_train_data, all_train_label, all_test_data, all_test_label, \
all_validate_data, all_validate_label = [], [], [], [], [], []
label = sio.loadmat(input_path + 'label.mat')['label'][0]
i = 0
for _, _, c in os.walk(input_path):
for item in c:
if '_' in item and i <= use_person:
i += 1
# Prepare data
tmp_file = sio.loadmat(input_path + item, verify_compressed_data_integrity=False)
one_train_data, one_train_label, one_validate_data, one_validate_label, \
one_test_data, one_test_label = read_a_mat(tmp_file, label)
# print('One person Shapes:', one_train_data.shape, one_validate_data.shape, \
# one_test_data.shape)
all_train_data.append(one_train_data)
all_train_label.append(one_train_label)
all_validate_data.append(one_validate_data)
all_validate_label.append(one_validate_label)
all_test_data.append(one_test_data)
all_test_label.append(one_test_label)
all_train_data = vstack_list(all_train_data)
all_train_label = vstack_list(all_train_label)
all_validate_data = vstack_list(all_validate_data)
all_validate_label = vstack_list(all_validate_label)
all_test_data = vstack_list(all_test_data)
all_test_label = vstack_list(all_test_label)
# Shuffle
# all_train_data, all_train_label = shuffle_train(all_train_data, all_train_label)
return all_train_data, all_train_label, all_validate_data, \
all_validate_label, all_test_data, all_test_label