def pca_classification(physiological_data, labels, classes):
ppg_data = physiological_data[:, :, :]
print(physiological_data.shape, labels.shape, ppg_data.shape)
ppg_train, ppg_test, \
y_train, y_test = \
normal_train_test_split(ppg_data, labels)
scaler = StandardScaler()
# Fit on training set only.
scaler.fit(ppg_train
) # Apply transform to both the training set and the test set.
ppg_train = scaler.transform(ppg_train)
ppg_test = scaler.transform(ppg_test)
# Make an instance of the Model
pca = PCA(.95)
pca.fit(ppg_train)
ppg_train = pca.transform(ppg_train)
ppg_test = pca.transform(ppg_test)
preds_ppg = \
physiological_classification(ppg_train, ppg_test, y_train, y_test, classes)
print("ppg ppg ppg")
ppg_result = validate_predictions(preds_ppg, y_test, classes)
print(ppg_result)
def cnn_lstm_classification(physiological_data, labels, classes):
print(physiological_data.shape)
print(labels.shape)
train_x, test_x, \
train_y, test_y = \
normal_train_test_split(physiological_data, labels)
preds_physiological = cnn_lstm(train_x, test_x, train_y, test_y)
print(validate_predictions(preds_physiological, test_y, classes))
def cnn_lstm_classification(physiological_data, labels, classes):
print(physiological_data.shape)
print(labels.shape)
# train_x, test_x, train_y, test_y = \
# normal_train_test_split(physiological_data, labels)
# train_x, test_x, train_y, test_y = \
# leave_one_subject_out_split(physiological_data, labels, subject_out=0)
train_x, test_x, train_y, test_y = \
leave_one_trial_out_split(physiological_data, labels, trial_out=0)
preds_physiological = cnn_lstm(train_x, test_x, train_y, test_y)
print(validate_predictions(preds_physiological, test_y, classes))
def lstm_classification(physiological_data,
labels,
part_seconds,
classes,
sampling_rate=128):
'''
Classify data using lstm method
'''
participants, trials = np.array(labels).shape
all_physiological_features = []
print(physiological_data.shape)
participants, trials, points = physiological_data.shape
part_length = part_seconds * sampling_rate
part_count = int(points / part_length)
all_participants_labels = []
for p in range(participants):
all_trials_physiological = []
all_trial_labels = []
for t in range(trials):
physiological_parts = []
for i in range(part_count):
ppg_data = \
physiological_data[p,
t,
i * part_length:(i+1)*part_length]
ppg_labels = labels[p, t]
physiological_parts.append(get_ppg_features(ppg_data))
all_trial_labels.append(ppg_labels)
all_trials_physiological.append(physiological_parts)
all_participants_labels.append(all_trial_labels)
all_physiological_features.append(all_trials_physiological)
physiological_data = np.array(all_physiological_features)
all_participants_labels = np.array(all_participants_labels)
physiological_train, physiological_test, \
y_train, y_test = \
normal_train_test_split(physiological_data, all_participants_labels)
#physiological_train, physiological_test, \
# y_train, y_test = \
# leave_one_trial_out_split(physiological_data,
# np.array(all_participants_labels), trial_out=0)
preds_physiological = \
physiological_lstm_classifier(physiological_train, physiological_test,
y_train, y_test, classes)
print(validate_predictions(preds_physiological, y_test, classes))
def feature_classification(physiological_data,
labels,
part_seconds,
classes,
sampling_rate=128):
# physiological_data, labels = \
# prepare_experimental_data(classes, label_type, sampling_rate, ignore_time)
participants, trials = np.array(labels).shape
participants, trials, points = physiological_data.shape
all_physiological_features = []
part_length = part_seconds * sampling_rate
part_count = int(points / part_length)
all_participants_labels = []
for p in range(participants):
all_trials_physiological = []
all_trial_labels = []
for t in range(trials):
physiological_parts = []
all_parts_labels = []
for i in range(part_count):
physiological_parts.append(
get_gsr_features(
physiological_data[p, t, i * part_length:(i + 1) *
part_length]))
all_parts_labels.append(labels[p, t])
all_trial_labels.append(all_parts_labels)
all_trials_physiological.append(physiological_parts)
all_participants_labels.append(all_trial_labels)
all_physiological_features.append(all_trials_physiological)
physiological_data = np.array(all_physiological_features)
all_participants_labels = np.array(all_participants_labels)
physiological_train, physiological_test, \
y_train, y_test = \
leave_one_subject_out_split(physiological_data, all_participants_labels)
#physiological_train, physiological_test, \
# y_train, y_test = \
# leave_one_trial_out_split(physiological_data,
# np.array(all_participants_labels), trial_out=0)
preds_physiological = \
physiological_classification(
physiological_train, physiological_test, y_train, y_test, classes)
print(validate_predictions(preds_physiological, y_test, classes))