def bci_competition_chain(mode, avg_size):
if mode == 'B':
if avg_size == 5:
ch = [
my_tools.ReadSignal(files=get_files('bci_competition', 'B')),
my_tools.Montage(montage_type='no_montage'),
my_tools.LeaveChannels(channels=['C3']),
my_tools.Filter(
**{
'wp': 0.3,
'ws': 0.1,
'gpass': 3.0,
'gstop': 30.0,
'ftype': 'cheby2',
'unit': 'hz'
}),
my_tools.Filter(
**{
'wp': 15.0,
'ws': 25.0,
'gpass': 1.0,
'gstop': 60.0,
'ftype': 'ellip',
'unit': 'hz'
}),
my_tools.Segment(classes=('target', 'non-target'),
start_offset=0.15,
duration=0.6),
my_tools.Average(bin_selectors=[
lambda mgr: mgr['name'] == 'target',
lambda mgr: mgr['name'] == 'non-target'
],
bin_names=['target', 'non-target'],
size=avg_size,
baseline=0.2,
strategy='random'),
my_tools.Downsample(factor=7),
my_tools.Normalize(norm=2),
my_tools.PrepareTrainSet(),
my_tools.SVM(C=2.0,
Cmode='classProb',
kernel=ker.Polynomial(2))
]
return ch
elif avg_size == 10:
ch = [
my_tools.ReadSignal(files=get_files('bci_competition', 'B')),
my_tools.Montage(montage_type='no_montage'),
my_tools.LeaveChannels(channels=['Cz']),
my_tools.Filter(
**{
'wp': 0.3,
'ws': 0.1,
'gpass': 3.0,
'gstop': 30.0,
'ftype': 'cheby2',
'unit': 'hz'
}),
my_tools.Filter(
**{
'wp': 15.0,
'ws': 25.0,
'gpass': 1.0,
'gstop': 60.0,
'ftype': 'ellip',
'unit': 'hz'
}),
my_tools.Segment(classes=('target', 'non-target'),
start_offset=-0.2,
duration=0.6),
my_tools.Average(bin_selectors=[
lambda mgr: mgr['name'] == 'target',
lambda mgr: mgr['name'] == 'non-target'
],
bin_names=['target', 'non-target'],
size=avg_size,
baseline=0.0,
strategy='random'),
my_tools.Downsample(factor=9),
my_tools.Normalize(norm=2),
my_tools.PrepareTrainSet(),
my_tools.SVM(C=0.1,
Cmode='classProb',
kernel=ker.Polynomial(3))
]
return ch
elif avg_size == 15:
ch = [
my_tools.ReadSignal(files=get_files('bci_competition', 'B')),
my_tools.Montage(montage_type='no_montage'),
my_tools.LeaveChannels(channels=['Cz']),
my_tools.Filter(
**{
'wp': 0.3,
'ws': 0.1,
'gpass': 3.0,
'gstop': 30.0,
'ftype': 'cheby2',
'unit': 'hz'
}),
my_tools.Filter(
**{
'wp': 15.0,
'ws': 25.0,
'gpass': 1.0,
'gstop': 60.0,
'ftype': 'ellip',
'unit': 'hz'
}),
my_tools.Segment(classes=('target', 'non-target'),
start_offset=0.1,
duration=0.6),
my_tools.Average(bin_selectors=[
lambda mgr: mgr['name'] == 'target',
lambda mgr: mgr['name'] == 'non-target'
],
bin_names=['target', 'non-target'],
size=avg_size,
baseline=0.0,
strategy='random'),
my_tools.Downsample(factor=9),
my_tools.Normalize(norm=2),
my_tools.PrepareTrainSet(),
my_tools.SVM(C=12.0, Cmode='classProb', kernel=ker.Linear())
]
return ch
def run_plot(person1, mode1, person2, mode2, avg_sizes, channels):
chain1 = p300_sample_data.get_chain(person1, mode1, 5)
chain2 = p300_sample_data.get_chain(person2, mode2, 5)
x = 1 # for now
y = 1 # for now
assert(len(channels) == 1) # for now
mgrs1 = None
mgrs2 = None
# Filter
for ch in [chain1[0], chain1[3], chain1[4]]:
mgrs1 = ch.process(mgrs1)
for ch in [chain2[0], chain2[3], chain2[4]]:
mgrs2 = ch.process(mgrs2)
# Segment
segmenter = my_tools.Segment(classes=('target','non-target'), start_offset=-0.1, duration=0.6)
mgrs1 = segmenter.process(mgrs1)
mgrs2 = segmenter.process(mgrs2)
averager = my_tools.Average(bin_selectors=[lambda mgr: mgr['name'] == 'target',
lambda mgr: mgr['name'] == 'non-target'],
bin_names=['target', 'non-target'], size=5,
baseline=0.1, strategy='random')
samp = float(mgrs1[0].get_param('sampling_frequency'))
for i, avg_size in enumerate(avg_sizes):
# Ommit reading-in : normalize, prepare, classify
averager.size = avg_size
avgs1 = averager.process(mgrs1)
avgs2 = averager.process(mgrs2)
plotter = erp_plot.Plotter(x, y, avg_size)
plt_id = 0
for i_ch_name in channels:
plt_id += 1
avgs = []
for avg in avgs1:
if avg.get_param('epoch_name') == 'target':
avg.name = '1 - target'
avgs.append(avg)
break
for avg in avgs1:
if avg.get_param('epoch_name') == 'non-target':
avg.name = '1 - non-target'
avgs.append(avg)
break
for avg in avgs2:
if avg.get_param('epoch_name') == 'target':
avg.name = '2 - target'
avgs.append(avg)
break
for avg in avgs2:
if avg.get_param('epoch_name') == 'non-target':
avg.name = '2 - non-target'
avgs.append(avg)
break
for j, avg in enumerate(avgs):
ch = avg.get_channel_samples(i_ch_name)*0.0715
plotter.add_plot(ch,
avg.name+" - "+i_ch_name,
[i*(1/samp)*1000 for i in range(-0.1*samp-1, len(ch)-(0.1*samp))],
plt_id, {'linewidth':2.0})
plotter.prepare_to_show()
erp_plot.show()