def artifactify(dataset, window_size, randomly_add_artifacts=True):
dataset = dataset.clone()
artifact_list = []
artifact_dataset = DataSet()
spike_size = (window_size // 4) * 3
for window in ExperimentorService.windows(dataset, window_size):
artificer = Artificer(window)
if randomly_add_artifacts:
decision = random.randrange(0, 10)
if decision==1:
artifact_window = artificer.add_artifacts(spike_size)
artifact_list += [1]
else:
artifact_window = window
artifact_list += [0]
else:
artifact_window = artificer.add_artifacts(spike_size)
artifact_list += [1]
artifact_dataset += artifact_window
if len(dataset) > len(artifact_dataset):
amount_missing = len(dataset) - len(artifact_dataset)
return DataSet(artifact_dataset + dataset[-amount_missing:]), artifact_list
return artifact_dataset, artifact_list
def pca_reconstruction(dataset, window_size, threshold):
dataset = dataset.clone()
reconstructed_dataset = DataSet()
rejections = []
for window in ExperimentorService.windows(dataset, window_size):
artificer = Artificer(window)
reconstructed_window, _, _, rejected = artificer.pca_reconstruction(threshold=threshold)
reconstructed_dataset += reconstructed_window
rejections += [rejected]
return reconstructed_dataset, rejections
def calibrate(dataset, window_size=40):
dataset = dataset.clone()
threshold_max = 0
threshold_avg = 0
threshold_avg_max = []
for window in ExperimentorService.windows(dataset, window_size):
artificer = Artificer(window)
_, avg_eigenvalue, max_eigenvalue, _ = artificer.pca_reconstruction()
threshold_max = max(threshold_max, max_eigenvalue)
threshold_avg = max(threshold_avg, avg_eigenvalue)
threshold_avg_max += [max_eigenvalue]
threshold_avg_max = np.mean(threshold_avg_max)
return threshold_max, threshold_avg, threshold_avg_max
def test_knn_threshold(self):
# filename = '../../data/subject1_csv/eeg_200605191428_epochs/small.csv'
filename = '../../data/emotiv/EEG_Data_filtered.csv'
dataset = DataReader.read_data(filename, ',')
dataset_slice = DataSet(dataset[0:40])
artificer = Artificer(dataset_slice, add_artifacts=True)
threshold = artificer.knn_threshold()
artificer.pca_reconstruction(threshold)
artificer.visualize()
def test_pca(self):
# filename = '../../data/subject1_csv/eeg_200605191428_epochs/small.csv'
filename = '../../data/emotiv/EEG_Data_filtered.csv'
dataset = DataReader.read_data(filename, ',')
threshold = 0
for idx in range(len(dataset) // 40):
current_dataset = DataSet(dataset[idx * 40:(idx + 1) * 40])
if idx < 10:
artificer = Artificer(current_dataset, add_artifacts=False)
max_eigenvalue, rejected = artificer.pca_reconstruction()
threshold = max(threshold, max_eigenvalue)
else:
artificer = Artificer(current_dataset, add_artifacts=True)
artificer.pca_reconstruction(threshold)
artificer.visualize()
break
def test_pca_speed(self):
filename = '../../data/subject1_csv/eeg_200605191428_epochs/eeg_200605191428_epochs.csv'
dataset = DataReader.read_data(filename, ',')
print 'Length of dataset: ' + str(len(dataset))
print 'Number of windows: ' + str(len(dataset) // 40)
print 'Calculating thresholds...'
max_threshold = 0
avg_threshold = 0
for idx in range(25):
current_dataset = DataSet(dataset[idx * 40:(idx + 1) * 40])
artificer = Artificer(current_dataset, add_artifacts=False)
avg_eigenvalue, max_eigenvalue, rejected = artificer.pca_reconstruction()
max_threshold = max(max_threshold, max_eigenvalue)
avg_threshold = max(avg_threshold, avg_eigenvalue)
print 'Largest threshold: ' + str(max_threshold)
print 'Average threshold: ' + str(avg_threshold)
print '\nTiming PCA reconstructor...'
times = []
for idx in range(len(dataset) // 40):
current_dataset = DataSet(dataset[idx * 40:(idx + 1) * 40])
start_time_constructor = time.time()
artificer = Artificer(current_dataset, add_artifacts=False)
midway_time = time.time() - start_time_constructor
artificer.put_artifacts()
start_time_reconstructor = time.time()
artificer.pca_reconstruction(max_threshold)
times.append(time.time() - start_time_reconstructor + midway_time)
print 'Time per PCA reconstruction: ' + str(reduce(lambda x, y: x + y, times) / len(times)) + ' sec'
