def filter_epoching(raw,
freqs,
filt_transwidth,
baseline,
t_win,
resamp_rate=None,
mag=False):
"""Filter and epoch data for ERP analysis.
Filter, epoch and eventually downsample data.
NOTE: This function has values specific to the experiment hard-coded,
adjust before using for something else!!!
Parameters:
-----------
raw : Raw
raw data of one run.
freqs : tuple or list (len=2)
lower and upper limit for filtering (successive HP and LP filter)
filt_transwidth : tuple or list (len=2)
transition bandwidth for FIR filters (HP and LP).
baseline : tuple (len=2)
baseline limits.
t_win : tuple or list (len=2)
time window for epoching.
resamp_rate : None | float
resampling rate, if not None, data will be resampled
mag : bool
whether only magnetometers should be considered, defaults to False.
Returns
-------
events :
events used for epoching.
epochs_filtered : Epochs
filtered epochs
"""
# event business comes first - allows to get rid of stim channel later
# TODO: make this smarter and easier to control
events = find_events(raw, stim_channel='STI101', shortest_event=1)
sel = np.where(events[:, 2] <= 255)[0]
events = events[sel, :]
# Compensate for delay (as measured manually with photodiode)
events[:, 0] += int(.050 * raw.info['sfreq'])
# pick channels:
if mag is True:
raw.pick_types(meg='mag', eog=False, stim=False)
else:
raw.pick_types(meg=True, eog=False, stim=False)
# filtering
(str(kk) for kk in freqs) # handle the possibility of None type
print('Filtering from %s to %s Hz.' % (freqs[0], freqs[1]))
raw.filter(freqs[0],
freqs[1],
n_jobs=1,
l_trans_bandwidth=filt_transwidth[0],
h_trans_bandwidth=filt_transwidth[1],
fir_design='firwin')
# epoching
print('Epoching data.')
epochs_filtered = Epochs(raw,
events,
proj=False,
tmin=t_win[0],
tmax=t_win[1],
baseline=baseline,
preload=True)
epochs_filtered._raw = None # memory leakage
if resamp_rate is not None:
print('Resampling data at %i Hz.' % resamp_rate)
epochs_filtered.resample(resamp_rate)
return events, epochs_filtered
def hilbert_epoching(raw,
freqs,
filt_transwidth,
baseline,
t_win,
resamp_rate,
mag=False):
"""Get data ready for hilbert beamformer.
Filter, hilbertize, and epoch the data. Returns BP-filtered as well as
hilbertized data.
NOTE: This function has values specific to the experiment hard-coded,
adjust before using for something else!!!
Parameters:
-----------
raw : Raw
raw data of one run.
freqs : tuple or list (len=2)
lower and upper limit for BP-filtering.
filt_transwidth : float
transition bandwidth for FIR filter.
baseline : tuple (len=2)
baseline limits.
t_win : tuple or list (len=2)
time window for epoching.
resamp_rate : float
resampling rate, only hilbertized data will be resampled.
mag : bool
whether only magnetometers should be considered, defaults to False.
Returns
-------
events :
events used for epoching.
epochs_filtered : Epochs
bandpass-filtered epochs for covariance matrix estimation.
epochs_hilbert : Epochs
hilbertized epochs, resampled!.
"""
# event business comes first - allows to get rid of stim channel later
# TODO: make this smarter and easier to control
events = find_events(raw, stim_channel='STI101', shortest_event=1)
sel = np.where(events[:, 2] <= 255)[0]
events = events[sel, :]
# Compensate for delay (as measured manually with photodiode)
events[:, 0] += int(.050 * raw.info['sfreq'])
# pick channels:
if mag is True:
raw.pick_types(meg='mag', eog=False, stim=False)
else:
raw.pick_types(meg=True, eog=False, stim=False)
# filtering and hilbertizing
print('Bandpass-filtering and hilbertizing for %i to %i Hz.' %
(freqs[0], freqs[1]))
raw.filter(freqs[0],
freqs[1],
n_jobs=1,
l_trans_bandwidth=filt_transwidth,
h_trans_bandwidth=filt_transwidth,
fir_design='firwin')
raw_hilbert = raw.copy().apply_hilbert(n_jobs=1, envelope=False)
# epoching
print('Epoching data.')
epochs_filtered = Epochs(raw,
events,
proj=False,
tmin=t_win[0],
tmax=t_win[1],
baseline=baseline,
preload=True)
epochs_filtered._raw = None # memory leakage
epochs_hilbert = Epochs(raw_hilbert,
events,
proj=False,
tmin=t_win[0],
tmax=t_win[1],
baseline=baseline,
preload=True).resample(resamp_rate)
epochs_hilbert._raw = None # memory leakage
return events, epochs_filtered, epochs_hilbert
