forked from baanko/Imagined-Speech-protocol
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathread_open.py
228 lines (205 loc) · 8.52 KB
/
read_open.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
import mne
import numpy as np
from scipy import signal
from scipy.signal import butter, lfilter
from sklearn.preprocessing import normalize
def butter_bandpass(lowcut, highcut, fs, order=5):
nyq = 0.5 * fs
low = lowcut / nyq
high = highcut / nyq
b, a = butter(order, [low, high], btype='band')
return b, a
def butter_bandpass_filter(data, lowcut, highcut, fs, order=5):
b, a = butter_bandpass(lowcut, highcut, fs, order=order)
y = lfilter(b, a, data, axis=0)
return y
def read_file(eeg_fname, time_fname, filter=(0.1, 100), r=False):
ch_names = ['FT9', 'TP9', 'FT10', 'TP10', 'T7', 'T8']
info = mne.create_info(ch_names=ch_names, sfreq=250, ch_types=['eeg', 'eeg', 'eeg', 'eeg', 'eeg', 'eeg'])
eeg_file = open(eeg_fname, "r")
time_file = open(time_fname, "r")
target = ["3", "13", "23", "33", "43"]
time_stamp = 0
x = []
y = []
b, a = signal.iirnotch(60, 20, fs=250)
# b1, a1 = signal.iirnotch(50, 30, fs=250)
t_stamp = []
raw = []
for line in eeg_file.readlines():
if len(line.split()) > 1:
# c = []
# c.append(line.split()[1])
# c.extend(line.split()[3:])
c = line.split()[1:]
raw.append([float(i) for i in c])
else:
t_stamp.append(float(line))
# print(np.shape(raw))
raw = signal.lfilter(b, a, raw, axis=0)
# raw = signal.lfilter(b, a, raw, axis=0)
# raw = signal.lfilter(b1, a1, raw, axis=0)
raw = butter_bandpass_filter(raw, filter[0], filter[1], 250)
# print(np.shape(raw))
if r:
return mne.io.RawArray(np.transpose(raw), info)
c = 0
for line in time_file.readlines():
line = line.rstrip('\n')
time = line.split(" ,")[0]
label = line.split(" ,")[1]
if label in target:
while t_stamp[c] < float(time):
c += 1
data = []
for j in range(500):
data.append(raw[c-25])
c += 1
x.append(np.transpose(data))
# if target.index(label) > 1:
# y.append(1)
# else:
# y.append(0)
# if label == "3":
# y.append(0)
# else:
# y.append(1)
y.append(target.index(label))
# print(np.shape(x))
events = [range(len(y)), [0]*len(y), y]
events = np.asarray(events)
events = events.T
# print(np.shape(events))
epoch = mne.EpochsArray(x, info=info, events=events, event_id={target[0]: 0, target[1]: 1, target[2]: 2, target[3]: 3, target[4]: 4})
epoch = epoch.apply_baseline(baseline=(0, 0.1))
scale = mne.decoding.Scaler(epoch.info)
x = scale.fit_transform(epoch.get_data())
# print(x[0][0])
epoch = mne.EpochsArray(x, info=info, events=events, event_id={target[0]: 0, target[1]: 1, target[2]: 2, target[3]: 3, target[4]: 4})
montage = mne.channels.make_standard_montage('standard_1005')
epoch.set_montage(montage)
# epochs.filter(60, 50)
return epoch
# rms, zero crossing rate, moving window average, kurtosis, power spectral entropy
#n magnitude of FFT, discrete time wavelet based spectral entropy (db4)
# power spectral entrophy (delta, theta, alpha and beta), hurst exponent, petrosian fractal dimension
def compute_spectral(signal, sampling_rate, bands=None):
psd = np.abs(np.fft.rfft(signal)) ** 2
psd /= np.sum(psd) # psd as a pdf (normalised to one)
if bands is None:
power_per_band = psd[psd > 0]
else:
freqs = np.fft.rfftfreq(signal.size, 1 / float(sampling_rate))
bands = np.asarray(bands)
freq_limits_low = np.concatenate([[0.0], bands])
freq_limits_up = np.concatenate([bands, [np.Inf]])
power_per_band = [np.sum(psd[np.bitwise_and(freqs >= low, freqs < up)])
for low, up in zip(freq_limits_low, freq_limits_up)]
power_per_band = np.array(power_per_band)[np.array(power_per_band) > 0]
spectral = - np.sum(power_per_band * np.log2(power_per_band))
return spectral
import librosa
import scipy
import math
def extract_feature_set_one(epoch, shape='ect'):
feature_set = []
sr = 250
if np.shape(epoch)[2] > 600:
sr = 500
for i in range(np.shape(epoch)[0]):
features = []
for j in range(int(np.shape(epoch)[2] / (sr/50))):
ch = []
for k in range(np.shape(epoch)[1]):
e = epoch[i][k][j*int(sr/50):(j+1)*int(sr/50)]
# zero_crossing = librosa.feature.zero_crossing_rate(epoch[i][k], frame_length=len(epoch[i][k]))
# print(e)
# print('zc')
zero_crosses = np.nonzero(np.diff(e > 0))[0]
zero_crossing = zero_crosses.size
# print('kt')
kurtosis = scipy.stats.kurtosis(e)
if math.isnan(kurtosis):
kurtosis = 0
# k1 = scipy.stats.kurtosis(epoch[0][1])
# print(np.shape(kurtosis))
# print(k1)
# print(kurtosis[0][1])
# print('spectral')
spectral = compute_spectral(e, sr)
# print('rms')
rms = np.sqrt(np.mean(np.square(e)))
mean = np.mean(e)
feature = [zero_crossing, kurtosis, spectral, rms, mean]
# print("epoch")
# print(e)
# print("feature")
# print(feature)
ch.extend(feature)
features.append(ch)
# print(np.shape(features))
if shape == 'ect':
feature_set.append(np.transpose(features))
else:
feature_set.append(features)
# print(np.shape(feature_set))
return np.asarray(feature_set)
def read_band_features(eeg_fname, time_fname, filterband = [(0.1, 4),(4,8),(8,15),(15,32),(32,None)]):
ch_names = ['FT9', 'TP9', 'FT10', 'TP10', 'T7', 'T8']
info = mne.create_info(ch_names=ch_names, sfreq=250, ch_types=['eeg', 'eeg', 'eeg', 'eeg', 'eeg', 'eeg'])
eeg_file = open(eeg_fname, "r")
time_file = open(time_fname, "r")
target = ["3", "13", "23", "33", "43"]
time_stamp = 0
x = []
y = []
b, a = signal.iirnotch(60, 30, fs=250)
# b1, a1 = signal.iirnotch(50, 30, fs=250)
t_stamp = []
raw = []
for line in eeg_file.readlines():
if len(line.split()) > 1:
c = line.split()[1:]
raw.append([float(i) for i in c])
else:
t_stamp.append(float(line))
# print(np.shape(raw))
raw = signal.lfilter(b, a, raw, axis=0)
# raw = signal.lfilter(b1, a1, raw, axis=0)
# raw = butter_bandpass_filter(raw, filter[0], filter[1], 250)
fr = []
# r = mne.io.RawArray(np.transpose(raw), info)
for i in len(filterband):
fr.append(butter_bandpass_filter(raw, filterband[i][0], filterband[i][1], 250))
c = 0
e = [[]]*len(fr)
for line in time_file.readlines():
line = line.rstrip('\n')
time = line.split(" ,")[0]
label = line.split(" ,")[1]
if label in target:
while t_stamp[c] < float(time):
c += 1
for ch in len(fr):
data = []
for h in range(500):
data.append(fr[h][c])
c += 1
e[ch].append(np.transpose(data))
# x.append(np.transpose(data))
y.append(target.index(label))
events = [range(len(y)), [0]*len(y), y]
events = np.asarray(events)
events = events.T
epoch_list = []
for ch in len(fr):
epoch = mne.EpochsArray(e[ch], info=info, events=events, event_id={target[0]: 0, target[1]: 1, target[2]: 2, target[3]: 3, target[4]: 4})
# epoch = mne.EpochsArray(x, info=info, events=events, event_id={target[0]: 0, target[1]: 1, target[2]: 2, target[3]: 3, target[4]: 4})
scale = mne.decoding.Scaler(epoch.info)
x = scale.fit_transform(epoch.get_data())
epoch_list.append(mne.EpochsArray(x, info=info, events=events, event_id={target[0]: 0, target[1]: 1, target[2]: 2, target[3]: 3, target[4]: 4}))
return epoch_list
# print(complexity["Entropy_Spectral"])
# epoch = read_file("Ear/0515_SH01/0515_PRESTIM_eeg_time_50t5c2s32ch_SH01_1.txt", "Ear/0515_SH01/0515_PRESTIM_aligned_speech_10t5c2s32ch_SH01_1.txt")
# f = extract_feature_set_one(epoch.get_data())
# print(np.shape(f))