def create_initial_EEGs(raw_input):
data_in = raw_input[1:, 1:30]
tables_array = np.split(data_in, NO_OF_EEGS)
EEGs = []
for table in tables_array:
subject = str(table[0, 0])
class_nr = int(table[0, 2])
class_label = str(table[0, 3])
channels = []
for i in range(NO_OF_CHANNELS):
channel_label = str(table[i, 5])
MFCCs = table[i, 11:]
MFCC_list = []
for j in range(NO_OF_MFCCS):
MFCC_list.append(float(MFCCs[j]))
channel = Channel(i, channel_label, MFCC_list)
channels.append(channel)
EEGs.append(EEG(subject, class_nr, class_label, channels))
return EEGs
def trials(trial_name=''):
error_master = []
no_error_master = []
trial_dicts = {
'Offline': offline_dict,
'S2': S2_dict,
'S3': S3_dict,
}
# Create or load the data
if not (os.path.exists(f'{trial_name}_error.npy') and os.path.exists(f'{trial_name}_no_error.npy')):
trial = trial_dicts[trial_name]
for subject in trial.keys():
runs = trial[subject]
# count = 0
for run in runs:
path = os.path.join(root, subject, trial_name, run)
e = EEG()
(eeg, trig, mat) = e.open(
path,
channel_names=channel_names_spellers
)
#theta = e.getTheta()
theta = e.getRawEEG(channels=channels_of_interest)[0]
theta = e.CARFilter(theta)
powerFeature = PowerFeature()
theta, grand_avg, gran_var = powerFeature.extract(theta,
window=32,
overlap=16
)
# theta = np.expand_dims(theta, axis=1)
error = e.splitNDArray(theta, 0, 1, True, fs=512, downsample=16)
no_error = e.splitNDArray(theta, 0, 1, False, fs=512, downsample=16)
if error_master == []:
error_master = error
no_error_master = no_error
else:
error_master = np.append(error_master, error, axis=0)
no_error_master = np.append(no_error_master, no_error, axis=0)
np.save(f'{trial_name}_error', error_master)
np.save(f'{trial_name}_no_error', no_error_master)
else:
error_master = np.load(f'{trial_name}_error.npy')
no_error_master = np.load(f'{trial_name}_no_error.npy')
#error_master = error_master[..., 1:]
error_master = np.swapaxes(error_master, 1, 2)
#no_error_master = no_error_master[..., 1:]
no_error_master = np.swapaxes(no_error_master, 1, 2)
error_master *= 10e6 # zscore(error_master)
no_error_master *= 10e6 # zscore(no_error_master)
return error_master, no_error_master
def __init__(self, amt_trials):
print(self.SHAPE)
self.robot = Robot.Robot.eeg_side_quickstart(self.URI)
self.distr = Distributor()
self.eeg = EEG.EEG(self.PATH)
self.net = Net.Net(self.SHAPE)
self.labels = []
self.datalist = []
self.eventlist = []
self.amt_trials = amt_trials # Amount of trials
def trials(trial_name=''):
error_master = []
no_error_master = []
trial_dicts = {
'Offline': offline_dict,
}
# Create or load the data
if not (os.path.exists(f'{trial_name}_error.npy')
and os.path.exists(f'{trial_name}_no_error.npy')):
trial = trial_dicts[trial_name]
for subject in trial.keys():
runs = trial[subject]
# count = 0
for run in runs:
path = os.path.join(root, subject, trial_name, run)
e = EEG()
(eeg, trig, mat) = e.open(path,
channel_names=channel_names_spellers)
theta = e.getRawEEG(channels=channels_of_interest)[0]
theta = e.CARFilter(theta)
stftfeature = STFTFeature()
stft, freq, time = stftfeature.extract(theta,
window=512,
overlap=496,
pre_trigger_time=0,
fs=512,
frequency_range=[3, 9])
error = e.splitNDArray(stft,
0,
1,
True,
fs=512,
downsample=(512 - 496))
no_error = e.splitNDArray(stft,
0,
1,
False,
fs=512,
downsample=(512 - 496))
if error_master == []:
error_master = error
no_error_master = no_error
else:
error_master = np.append(error_master, error, axis=0)
no_error_master = np.append(no_error_master,
no_error,
axis=0)
np.save(f'{trial_name}_error', error_master)
np.save(f'{trial_name}_no_error', no_error_master)
else:
error_master = np.load(f'{trial_name}_error.npy')
no_error_master = np.load(f'{trial_name}_no_error.npy')
error_master = np.swapaxes(error_master, 2, 3)
error_master = np.swapaxes(error_master, 1, 3)
no_error_master = np.swapaxes(no_error_master, 2, 3)
no_error_master = np.swapaxes(no_error_master, 1, 3)
error_master *= 10e6 # zscore(error_master)
no_error_master *= 10e6 # zscore(no_error_master)
return error_master, no_error_master
feature_vol, grand_avg, grand_var = feature.extract(
error, window, overlap)
return feature_vol, grand_avg, grand_var
for subject in offline_dict.keys():
runs = offline_dict[subject]
grand_avg_fig, grand_avg_axis = plt.subplots(len(channel_names_spellers),
3)
count = 0
for run in runs:
path = os.path.join(root, subject, 'Offline', run)
e = EEG()
(eeg, trig, mat) = e.open(path, channel_names=channel_names_spellers)
#theta = e.getTheta()
theta = eeg
error = e.getEEGTrials(
pretime=.5,
posttime=1,
error=True,
plot=False,
gdf=theta,
)
no_error = e.getEEGTrials(pretime=.5,
posttime=1,
error=False,