/
SSVEP1.py
374 lines (294 loc) · 14.6 KB
/
SSVEP1.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
#Required libraries
import mne
from mne import Epochs,find_events
import numpy as np
import matplotlib.pyplot as plt
import scipy
import os
from collections import OrderedDict
import seaborn as sns
import pandas as pd
import gzip
from scipy.signal import filtfilt, butter
import pickle
from pyriemann.utils.distance import distance_riemann
#from pyriemann.utils.mean import mean_riemann
#from pyriemann.tangentspace import TangentSpace
from estimation import covariances
from riemannian_geometry import mean_riemann,project
from itertools import combinations,product
from tqdm import tqdm
from sklearn.linear_model import LogisticRegression
from sklearn.base import BaseEstimator, TransformerMixin
from sklearn.covariance import shrunk_covariance
def SsvepLoading(data_path):
"""
Inputs:
data_path: path for the ssvep recordings
Outputs:
subj_list : a list of subjects ["subject1",...]
records : a dictionnary of subjects and their associated sessions
"""
subj_list = os.listdir(data_path)
records = {s: [] for s in range(len(subj_list))}
for s in range(len(subj_list)):
subj = subj_list[s]
record_all = os.listdir(data_path+subj+'/')
for file in record_all:
if file[len(file)-8:]=="_raw.fif":
records[s].append(file[:len(file)-8])
return subj_list,records
def filter_bandpass(signal, lowcut, highcut, fs, order=4, filttype='forward-backward'):
nyq = 0.5 * fs
low = lowcut / nyq
high = highcut / nyq
b, a = butter(order, [low, high], btype='band')
if filttype == 'forward':
filtered = lfilter(b, a, signal, axis=-1)
elif filttype == 'forward-backward':
filtered = filtfilt(b, a, signal, axis=-1)
else:
raise ValueError("Unknown filttype:", filttype)
return filtered
class TrialsBuilding():
def __init__(self,data_path,records,subj_list,subject,nb_classes,tmin,tmax,freq_band):
self.records = records
self.data_path = data_path
self.subj_list = subj_list
self.subject = subject
self.nb_classes = nb_classes
assert 0<= self.subject <len(self.subj_list),"The selected subject does not exist in the dataset"
self.nb_total_sessions = len(self.records[self.subject])
#experimental setup
self.tmin=tmin
self.tmax=tmax
self.sfreq = 256
self.freq_band=freq_band
self.frequencies= [13,17,21]
if self.nb_classes==4:
self.event_code = [33024,33025,33026,33027]
self.event_code_fif = [1,2,3,4]
#self.names=['resting','stim13','stim21','stim17']
else:
self.event_code = [33025,33026,33027]
self.event_code_fif = [2,3,4]
#self.names=['stim13','stim21','stim17']
self.channels = np.array(['Oz','O1','O2','PO3','POz','PO7','PO8','PO4'])
def load_single_session(self,subject,session):
chosen_subject = self.subj_list[subject]
fname = chosen_subject+'/'+self.records[subject][session]
if os.path.exists(self.data_path + fname + '.pz'):
with gzip.open(self.data_path + fname + '.pz', 'rb') as f:
o = pickle.load(f, encoding='latin1')
raw_signal = o['raw_signal'].T
event_pos = o['event_pos'].reshape((o['event_pos'].shape[0]))
event_type = o['event_type'].reshape((o['event_type'].shape[0]))
data_type = "pz"
else:
raw = mne.io.read_raw_fif(self.data_path + fname + '_raw.fif',preload=True)
raw_signal = raw.get_data()
raw_signal = raw_signal[:raw_signal.shape[0]-1,:]
events = mne.find_events(raw)
event_pos = events.T[0]
event_type = events.T[2]
data_type = "fif"
return raw_signal,event_pos,event_type,data_type
def make_extended_trials_single_session(self,raw_signal,event_pos,event_type,data_type):
ext_signal = np.empty_like(raw_signal[0,:]) #(1,n)
for f in self.frequencies:
ext_signal = np.vstack((ext_signal, filter_bandpass(raw_signal, f-self.freq_band,
f+self.freq_band, fs=self.sfreq)))
ext_signal = ext_signal[1:,:]
ext_trials = list()
for i in range(len(event_type)):
if data_type=="pz":
boolean = (event_type[i] == 32779) and (i>0) and (event_type[i-1] in self.event_code) # start of a trial
if data_type=="fif":
boolean = event_type[i] in self.event_code_fif
if boolean:
t = event_pos[i]
start = int(t + self.tmin*self.sfreq)
stop = int(t + self.tmax*self.sfreq)
ext_trials.append(ext_signal[:, start:stop])
ext_trials = np.array(ext_trials)
ext_trials = ext_trials - np.tile(ext_trials.mean(axis=2).reshape(ext_trials.shape[0],
ext_trials.shape[1], 1), (1, 1, ext_trials.shape[2]))
return ext_trials
def make_labels_single_session(self,event_type,data_type):
labels = []
for e in event_type:
if data_type=="pz":
for i in range(self.nb_classes):
if e==self.event_code[i]:
labels.append(i)
if data_type=="fif":
if self.nb_classes==4:
event_idx={1:0,2:1,3:3,4:2}#resting = 1, 13Hz = 2, 21Hz = 3, 17Hz = 4
else:
event_idx={2:0,3:2,4:1}
if (e in self.event_code_fif):
labels.append(event_idx[e])
return labels
def extended_trials_and_labels_all_sessions(self):
ext_trials_all_sessions = []
labels_all_sessions = [] #list of length 32*total_nb_sessions
for session in range(len(self.records[self.subject])):
raw_signal,event_pos,event_type,data_type = self.load_single_session(self.subject,session)
ext_trial_single = self.make_extended_trials_single_session(raw_signal,event_pos,event_type,data_type)
labels = self.make_labels_single_session(event_type,data_type)
ext_trials_all_sessions.append(ext_trial_single)
labels_all_sessions.extend(labels)
n_trials_per_session,dim1,dim2 = ext_trial_single.shape #(32,24,768) or (24,24,768)
extended_trials = np.zeros((self.nb_total_sessions*n_trials_per_session,dim1,dim2))
#shape : (32*total_nb_sessions,24,768) if nb_clases=4
#shape : (24*total_nb_sessions,24,768) if nb_clases=3
for i in range(len(ext_trials_all_sessions)):
extended_trials[n_trials_per_session*i:n_trials_per_session*(i+1),:,:]= ext_trials_all_sessions[i]
ch = "There is a problem of shapes : "+str(len(labels_all_sessions))+" != "+str(extended_trials.shape[0])
assert len(labels_all_sessions)==extended_trials.shape[0],ch
return extended_trials,labels_all_sessions
class Classify():
def __init__(self,method,covs,labels,nb_trains,nb_classes,with_shuffle=False, train_prop = 0.75,kfold=10,u_prime=lambda x : 1):
self.covs = covs #all covs , for test and trainn
self.labels = labels
self.method = method
self.u_prime=u_prime
self.with_shuffle = with_shuffle
self.train_prop = train_prop
self.kfold = kfold
assert method in ["MDM","TangentSpace"]
self.nb_classes = nb_classes
self.nb_total_trials = covs.shape[0]
self.nb_total_sessions = self.nb_total_trials//(8*self.nb_classes)
self.nb_trains = nb_trains
#if nb_trains=0, we choose a session and we split into train and test sets
#if nb_trains=nb_total_sessions, we take nb_total_sessions-1 and split the last session into train and test
assert 0<= self.nb_trains <= self.nb_total_sessions,"Make sure that the nbr of training sets is <= to the nber of sessions"
def classifier(self, x_train, y_train ):
if self.method=="MDM":
return MDM(x_train,y_train,self.nb_classes,u_prime=self.u_prime)
if self.method=="TangentSpace":
return Tangent_Space(x_train,y_train,self.nb_classes,u_prime=self.u_prime)
def shuffle_sessions(self):
idx_per_class = { k : [] for k in range(self.nb_classes)}
for i in range(len(self.labels)):
idx_per_class[self.labels[i]].append(i)
for k in range(self.nb_classes):
np.random.shuffle(idx_per_class[k]) #length=8*nb_sessions
shuffled_idx = []
nb_samples_per_class = len(idx_per_class[0])//self.nb_total_sessions #=8
for n in range(self.nb_total_sessions):
session_idx= []
for k in range(self.nb_classes):
session_idx.extend(idx_per_class[k][n*nb_samples_per_class:(n+1)*nb_samples_per_class])
shuffled_idx.extend(session_idx)
return shuffled_idx
def split(self, with_shuffle=False,train_prop=0.75,kfold=5):
trains_idx , tests_idx = [],[]
sessions = list(range(self.nb_total_sessions))
if with_shuffle:
indices = self.shuffle_sessions()
else:
indices = list(range(len(self.labels)))
if (1<=self.nb_trains<self.nb_total_sessions):
test_sessions_idx = list(combinations(sessions,self.nb_total_sessions-self.nb_trains))
for i in range(len(test_sessions_idx)):
test_idx , train_idx = [],[]
test_session_idx = test_sessions_idx[i]
for j in test_session_idx:
test_idx.extend(indices[j*8*self.nb_classes:(j+1)*8*self.nb_classes])
for k in indices:
if not(k in test_idx):
train_idx.append(k)
trains_idx.append(train_idx)
tests_idx.append(test_idx)
if self.nb_trains ==0 : #then apply the split 0.75 for train and 0.25 for test + kfold cross validation
assert 0<train_prop<1
train_per_class = int(8*train_prop)
idx_per_class = { k : [] for k in range(self.nb_classes)}
for i in range(len(self.labels)):
idx_per_class[self.labels[i]].append(i)
for c in range(kfold):
test_idx,train_idx = [],[]
for k in range(self.nb_classes):
np.random.shuffle(idx_per_class[k])
train_idx.extend(idx_per_class[k][:self.nb_total_sessions*train_per_class])
test_idx.extend(idx_per_class[k][self.nb_total_sessions*train_per_class:])
trains_idx.append(train_idx)
tests_idx.append(test_idx)
return trains_idx,tests_idx
def listing(self,covs,labels,indices):
x,y= [],[]
for i in indices:
x.append(covs[i,:,:])
y.append(labels[i])
return x,y
def score(self, predicted_labels,true_labels):
assert len(predicted_labels)==len(true_labels),"True and predicted labels' lists haven't the same length"
acc = 0
for j in range(len(predicted_labels)):
if predicted_labels[j]==true_labels[j]:
acc +=1
return acc/len(predicted_labels)
def accuracies(self):
trains_idx,tests_idx = self.split(self.with_shuffle,self.train_prop,self.kfold)
accuracies_train = []
accuracies_test = []
for train_idx , test_idx in zip(trains_idx , tests_idx):
covs_train, true_labels_train = self.listing(self.covs,self.labels,train_idx)
classifier = self.classifier(covs_train, true_labels_train)
predicted_labels_train = classifier.predict(covs_train)
accuracy_train = self.score(predicted_labels_train,true_labels_train)
accuracies_train.append(accuracy_train)
covs_test, true_labels_test = self.listing(self.covs,self.labels,test_idx)
predicted_labels_test = classifier.predict(covs_test)
accuracy_test = self.score(predicted_labels_test,true_labels_test)
accuracies_test.append(accuracy_test)
return np.asarray(accuracies_train),np.asarray(accuracies_test)
class MDM():
def __init__(self,x_train,y_train,nb_classes, u_prime = lambda x : 1):
self.nb_classes = nb_classes
self.u_prime = u_prime
self.cov_centers = self.MassCenters(x_train,y_train)
def MassCenters(self,x_train,y_train):
cov_centers = np.empty((self.nb_classes, x_train[0].shape[1], x_train[0].shape[1]))
x_train_bis = np.empty((len(x_train),x_train[0].shape[1], x_train[0].shape[1]))
for i in range(len(x_train)):
x_train_bis[i,:,:] = x_train[i]
classes = list(range(self.nb_classes))
y_train_bis = np.asarray(y_train)
for i, l in enumerate(classes):
cov_centers[i, :, :] = mean_riemann(x_train_bis[y_train_bis==l,:,:],u_prime = self.u_prime) ######
return cov_centers
def argmin_distance(self,sample):
min_dist = np.inf
for i in range(self.nb_classes):
dist = distance_riemann(sample, self.cov_centers[i])
if min_dist > dist:
min_dist = dist
idx = i
return idx
def predict(self,x):
prediction = []
for sample in x:
predicted_label = self.argmin_distance(sample)
prediction.append(predicted_label)
return prediction
class Tangent_Space():
def __init__(self,x_train,y_train,nb_classes,reference=None,u_prime=lambda x:1):
self.nb_classes = nb_classes
self.reference = refence
self.clf = self.ProjectedClassifier(x_train,y_train)
self.u_prime = u_prime
def tangent_project(self,x_train,y_train):
if self.reference ==None:
self.reference = mean_riemann(np.asarray(x_train),u_prime = self.u_prime)
x_train_proj = project(self.reference,x_train)
return x_train_proj
def ProjectedClassifier(self,x_train,y_train):
x_train_proj = self.tangent_project(x_train,y_train)
clf = LogisticRegression(random_state=0).fit(x_train_proj,y_train)
return clf
def predict(self,x,y):
x_proj = self.tangent_project(x,y)
return self.clf.predict(x_proj)