def getonefeatures(data):
listdata = list(data)
# Std
fstd = np.array(list(map(lambda x: np.std(x), listdata)))
# Approximate-Entropy(ApEN)
m = 3
r = 0.2 * fstd
fae = np.array(
list(map(lambda x, y: pyeeg.ap_entropy(x, m, y), listdata, list(r))))
# Power
# fpower,fre = pyeeg.bin_power(data, [1,30], fs)
# print(fpower)
# print("特征--{std:%.4f,AE:%.4f,Power:%.4f}"%(fstd,fae,fpower))
# First-order Diff ???
# firstoderlist = pyeeg.first_order_diff(data)
# Hjorth
fhjormod_com = np.array(list(map(lambda x: pyeeg.hjorth(x), listdata)))
fhjor_act = np.array(list(map(lambda x: np.var(x), listdata)))
# Spectrum Entropy
# fse = pyeeg.spectral_entropy(data, [0,fs/2], fs)
# print(fse)
# Power Spectral Density
fpsd = np.array(list(map(lambda x: np.max(plt.psd(x, fs)[0]), listdata)))
# Features Stack
featurestmp = np.stack(
(fstd, fae, fhjormod_com[:, 0], fhjormod_com[:, 1], fhjor_act, fpsd),
axis=1)
temprow, tmpcol = featurestmp.shape
features = np.reshape(featurestmp, (temprow * tmpcol, ))
return features
def extract_features(couple_data):
pca1 = pca_project_data(couple_data, 1) #take 1st pca dimension
pca1_mean = np.mean(pca1, axis=0) #
pca1_std = np.std(pca1, axis=0) #
pca1_med = np.median(pca1, axis=0) #
features = []
def sinuosity_deviation_features(seq, mean, std):
sinuosity_dict = {"A":0, "B":0, "C":0}
deviation_dict = {"I":0, "II":0, "III":0}
sinuosity_deviation_dict = {"A-I":0,"A-II":0,"A-III":0,"B-I":0,"B-II":0,"B-III":0,"C-I":0,"C-II":0,"C-III":0}
n = len(seq)
for i in range(1,n-1):
current_af = seq[i]
prev_af = seq[i-1]
next_af = seq[i+1]
sinu = abs((next_af - current_af) + (current_af - prev_af))
if sinu == 0: label1 = "A"
elif 0< sinu <= 1: label1 = "B"
else: label1 = 'C'
sinuosity_dict[label1] += 1
devi = abs(current_af - mean)
close = std / 2
if devi <= close: label2 = "I"
elif devi <= std: label2 = "II"
elif devi > std: label2 = "III"
deviation_dict[label2] += 1
sinuosity_deviation_dict["%s-%s"%(label1,label2)] += 1
return sinuosity_deviation_dict.values()
n = len(pca1)
pca1_sinuosity_deviation = sinuosity_deviation_features( pca1, pca1_mean, pca1_std )
features += list(np.array(pca1_sinuosity_deviation)/float(n-2))
seq = pca1
dfa = eg.dfa(seq); pfd = eg.pfd(seq)
apen = eg.ap_entropy(seq,1,np.std(seq)*.2)
svden = eg.svd_entropy(seq, 2, 2)
features += [pca1_mean, pca1_med, pca1_std, dfa, pfd, apen, svden]
return features
MAX_EPOCH_N = 256 * 30
EPOCH_STEP = 256 * 5
N_REPLICATES = 5
SPECT_ENT_BANDS = 2 ** np.arange(0,8)/2
fun_to_test = [
{"times":100,"name":"hfd", "is_original":True,"fun": lambda x: pyeeg.hfd(x,2**3)},
{"times":100,"name":"hfd", "is_original":False,"fun": lambda x: univ.hfd(x,2**3)},
{"times":100,"name":"hjorth", "is_original":True,"fun": lambda x: pyeeg.hjorth(x)},
{"times":100,"name":"hjorth", "is_original":False,"fun": lambda x: univ.hjorth(x)},
{"times":100,"name":"pfd", "is_original":True, "fun":lambda x: pyeeg.pfd(x)},
{"times":100,"name":"pfd", "is_original":False, "fun":lambda x: pyeeg.pfd(x)},
{"times":2,"name":"samp_ent", "is_original":True, "fun":lambda x: pyeeg.samp_entropy(x,2,1.5)},
{"times":10,"name":"samp_ent", "is_original":False, "fun":lambda x: univ.samp_entropy(x,2,1.5,relative_r=False)},
{"times":2,"name":"ap_ent", "is_original":True, "fun":lambda x: pyeeg.ap_entropy(x,2,1.5)},
{"times":10,"name":"ap_ent", "is_original":False, "fun":lambda x: univ.ap_entropy(x,2,1.5)},
{"times":10,"name":"svd_ent", "is_original":True, "fun":lambda x: pyeeg.svd_entropy(x,2,3)},
{"times":100,"name":"svd_ent", "is_original":False, "fun":lambda x: univ.svd_entropy(x,2,3)},
{"times":10,"name":"fisher_info", "is_original":True, "fun":lambda x: pyeeg.fisher_info(x,2,3)},
{"times":100, "name":"fisher_info", "is_original":False, "fun":lambda x: univ.fisher_info(x,2,3)},
{"times":100,"name":"spectral_entropy", "is_original":True, "fun":lambda x: pyeeg.spectral_entropy(x,SPECT_ENT_BANDS,256)},
{"times":100, "name":"spectral_entropy", "is_original":False, "fun":lambda x: univ.spectral_entropy(x,256, SPECT_ENT_BANDS)},
]
def make_one_rep():
ldfs = []
for n in range(MIN_EPOCH_N, MAX_EPOCH_N + 1, EPOCH_STEP):
a = numpy.random.normal(size=n)
def CreateFeatureVector(collection, dbName, takeFirstMinutes):
limitSamples = SamplingRates[dbName] * 60 * takeFirstMinutes
lastBeat = -1
lastQ = -1
amplitudeSum = {
Constants.LabelQ: 0,
Constants.LabelR: 0,
Constants.LabelS: 0
}
labelsCounters = {
Constants.LabelQ: 0,
Constants.LabelR: 0,
Constants.LabelS: 0
}
amplitudesLists = {
Constants.LabelQ: list(),
Constants.LabelR: list(),
Constants.LabelS: list()
}
heartbeats = list()
valuesHistogram = dict()
sumQS = 0
countQS = 0
for entry in collection.find().sort(Constants.Time,
ASCENDING).limit(limitSamples):
label = entry[Constants.Label]
time = entry[Constants.Time]
value = entry[Constants.Value]
if label == Constants.LabelNone:
continue
amplitudeSum[label] += value
labelsCounters[label] += 1
amplitudesLists[label].append(value)
if value in valuesHistogram:
valuesHistogram[value] += 1
else:
valuesHistogram[value] = 1
if label == Constants.LabelR:
if lastBeat > 0:
heartbeats.append(time - lastBeat)
lastBeat = time
elif label == Constants.LabelQ:
lastQ = time
elif label == Constants.LabelS:
if lastQ > 0:
sumQS += time - lastQ
countQS += 1
lastQ = -1
averageQAmplitude = amplitudeSum[Constants.LabelQ] / float(
labelsCounters[Constants.LabelQ])
averageRAmplitude = amplitudeSum[Constants.LabelR] / float(
labelsCounters[Constants.LabelR])
averageSAmplitude = amplitudeSum[Constants.LabelS] / float(
labelsCounters[Constants.LabelS])
# Calculate the baseline by the most common value
baseline = max(valuesHistogram.iteritems(), key=operator.itemgetter(1))[0]
# Convert heartbeats length from a number of samples to actual time
normalizedHeartbeats = [
float(i) / SamplingRates[dbName] for i in heartbeats
]
heartbeatSTD = std(normalizedHeartbeats)
beatChanges = [
abs(float(x) - normalizedHeartbeats[i - 1]) /
normalizedHeartbeats[i - 1] for i, x in enumerate(normalizedHeartbeats)
][1:]
beatChangesCount = len(
[change for change in beatChanges if float(change) >= 0.1])
totalBeatsCount = len(beatChanges)
result = FeatureVector(
Database=dbName,
RecordNumber=collection.name,
AverageHeartbeat=60 / mean(normalizedHeartbeats),
IrregularBeatsPercent=float(beatChangesCount) / totalBeatsCount,
AverageBeatChange=mean(beatChanges),
Irregularity=ap_entropy(normalizedHeartbeats, 2, heartbeatSTD * 0.2),
QS=(sumQS / float(countQS)) * 1000 / SamplingRates[dbName],
QtoR=abs(averageQAmplitude - baseline) /
abs(averageRAmplitude - baseline),
StoR=abs(averageSAmplitude - baseline) /
abs(averageRAmplitude - baseline),
QSTD=std(amplitudesLists[Constants.LabelQ]),
RSTD=std(amplitudesLists[Constants.LabelR]),
SSTD=std(amplitudesLists[Constants.LabelS]))
return result
#print (get_all_wavelet[0][0])
temp_extrasi1 = []
temp_wavelet1 = [] # for getting temporary array for fiture extraction
hasil_extrasi_fitur = [] #getting all result in feature extraction
for x in range(len(get_all_wavelet)):
for y in range(len(get_all_wavelet[x])):
temp_extrasi1.append(dc.energi(get_all_wavelet[x][y]))
temp_extrasi1.append(
np.std(get_all_wavelet[x]
[y])) # stdmu salah cak ji. aku langsung ambil ae iki.
temp_extrasi1.append(dc.maximum(get_all_wavelet[x][y]))
temp_extrasi1.append(dc.mininum(get_all_wavelet[x][y]))
temp_extrasi1.append(
ap_entropy(get_all_wavelet[x][y],
len(get_all_wavelet[x][y]) / 5, temp_extrasi1[1] *
float(0.2))) # diisi dengan aproximate entropy
#param 1 itu datanya param ke dua itu panjangnya data yang ingin d potong , param ke tiga itu similarity. param ke dua dan ketiga diisi dengan mencoba coba
temp_wavelet1.append(temp_extrasi1)
temp_extrasi1 = []
hasil_extrasi_fitur.append(temp_wavelet1)
temp_wavelet1 = []
# hasil seluruh extrasi fitur ada di hasil_extrasi_fitur dengan data berukuran 3D
print len(hasil_extrasi_fitur)
if len(hasil_extrasi_fitur) > int(0):
print len(hasil_extrasi_fitur[0])
print hasil_extrasi_fitur[0][0][0]
print hasil_extrasi_fitur[0][0][1]
print hasil_extrasi_fitur[0][0][2]
'tol': table[idx, 2],
'score': table[idx, 0]
})
print(' Channel %i: optimal pair is' % (i), table[idx, 1], 'min_l',
table[idx, 2], 'tol --', 'auc roc is', table[idx, 0])
print('Tuning aproximate entropy ...')
subwindow_sizes = np.arange(1, train_banded.shape[1])
opt_sw_tol_apent = []
for i in range(train_banded.shape[2]):
table = []
for subwindow in subwindow_sizes:
for tol in tolerance_values:
entropy_list = []
for j in range(train_banded.shape[0]):
entropy_list.append(
pyeeg.ap_entropy(train_banded[j, :, i], subwindow,
tol))
entropy_list = np.array(entropy_list)
score = np.max([
roc_auc_score(y_true=train_y, y_score=entropy_list),
roc_auc_score(y_true=1 - train_y, y_score=entropy_list)
])
table.append([score, subwindow, tol])
table = np.array(table)
idx = np.argmax(table[:, 0])
opt_sw_tol_apent.append({
'channel': i,
'subwindow': table[idx, 1],
'tol': table[idx, 2],
'score': table[idx, 0]
})
print(' Channel %i: optimal pair is' % (i), table[idx, 1],
def CreateFeatureVector(collection, dbName, takeFirstMinutes):
limitSamples = SamplingRates[dbName] * 60 * takeFirstMinutes
lastBeat = -1
lastQ = -1
amplitudeSum = {Constants.LabelQ: 0, Constants.LabelR: 0, Constants.LabelS: 0}
labelsCounters = {Constants.LabelQ: 0, Constants.LabelR: 0, Constants.LabelS: 0}
amplitudesLists = {Constants.LabelQ: list(), Constants.LabelR: list(), Constants.LabelS: list()}
heartbeats = list()
valuesHistogram = dict()
sumQS = 0
countQS = 0
for entry in collection.find().sort(Constants.Time, ASCENDING).limit(limitSamples):
label = entry[Constants.Label]
time = entry[Constants.Time]
value = entry[Constants.Value]
if label == Constants.LabelNone:
continue
amplitudeSum[label] += value
labelsCounters[label] += 1
amplitudesLists[label].append(value)
if value in valuesHistogram:
valuesHistogram[value] += 1
else:
valuesHistogram[value] = 1
if label == Constants.LabelR:
if lastBeat > 0:
heartbeats.append(time - lastBeat)
lastBeat = time
elif label == Constants.LabelQ:
lastQ = time
elif label == Constants.LabelS:
if lastQ > 0:
sumQS += time - lastQ
countQS += 1
lastQ = -1
averageQAmplitude = amplitudeSum[Constants.LabelQ] / float(labelsCounters[Constants.LabelQ])
averageRAmplitude = amplitudeSum[Constants.LabelR] / float(labelsCounters[Constants.LabelR])
averageSAmplitude = amplitudeSum[Constants.LabelS] / float(labelsCounters[Constants.LabelS])
# Calculate the baseline by the most common value
baseline = max(valuesHistogram.iteritems(), key=operator.itemgetter(1))[0]
# Convert heartbeats length from a number of samples to actual time
normalizedHeartbeats = [ float(i) / SamplingRates[dbName] for i in heartbeats ]
heartbeatSTD = std(normalizedHeartbeats)
beatChanges = [abs(float(x) - normalizedHeartbeats[i-1])/normalizedHeartbeats[i-1] for i, x in enumerate(normalizedHeartbeats)][1:]
beatChangesCount = len([change for change in beatChanges if float(change) >= 0.1])
totalBeatsCount = len(beatChanges)
result = FeatureVector(
Database = dbName,
RecordNumber = collection.name,
AverageHeartbeat = 60 / mean(normalizedHeartbeats),
IrregularBeatsPercent = float(beatChangesCount) / totalBeatsCount,
AverageBeatChange = mean(beatChanges),
Irregularity = ap_entropy(normalizedHeartbeats, 2, heartbeatSTD*0.2),
QS = (sumQS / float(countQS)) * 1000 / SamplingRates[dbName],
QtoR = abs(averageQAmplitude - baseline) / abs(averageRAmplitude - baseline),
StoR = abs(averageSAmplitude - baseline) / abs(averageRAmplitude - baseline),
QSTD = std(amplitudesLists[Constants.LabelQ]),
RSTD = std(amplitudesLists[Constants.LabelR]),
SSTD = std(amplitudesLists[Constants.LabelS])
)
return result
"times": 2,
"name": "samp_ent",
"is_original": True,
"fun": lambda x: pyeeg.samp_entropy(x, 2, 1.5)
},
{
"times": 10,
"name": "samp_ent",
"is_original": False,
"fun": lambda x: univ.samp_entropy(x, 2, 1.5, relative_r=False)
},
{
"times": 2,
"name": "ap_ent",
"is_original": True,
"fun": lambda x: pyeeg.ap_entropy(x, 2, 1.5)
},
{
"times": 10,
"name": "ap_ent",
"is_original": False,
"fun": lambda x: univ.ap_entropy(x, 2, 1.5)
},
{
"times": 10,
"name": "svd_ent",
"is_original": True,
"fun": lambda x: pyeeg.svd_entropy(x, 2, 3)
},
{
"times": 100,