def segment_heuristically(self, sensors, reference_signal, label=None):
p = PeakAnalysis(self.view)
peaks = p.get_peaks(reference_signal)
timestamps = []
for peak_position in peaks:
timestamps.append(sensors[0].timestamp[peak_position])
if label:
label_timestamps = label.timestamp
labels = label.label
closer_labels = []
closer_timestamps = []
for i in range(0, len(label_timestamps)):
min_dist = label_timestamps[len(label_timestamps) -
1] - label_timestamps[0]
closer_label = None
closer_timestamp = None
for heuristic_timestamp in timestamps:
dist = sqrt(
np.square(label_timestamps[i] - heuristic_timestamp))
if dist < min_dist:
min_dist = dist
closer_label = labels[i]
closer_timestamp = heuristic_timestamp
closer_labels.append(closer_label)
closer_timestamps.append(closer_timestamp)
self.segment(sensors, closer_timestamps, closer_labels)
else:
self.segment(sensors, timestamps)
def segment_heuristically(self, sensors, reference_signal, label=None):
p = PeakAnalysis(self.view)
peaks = p.get_peaks(reference_signal)
timestamps = []
for peak_position in peaks:
timestamps.append(sensors[0].timestamp[peak_position])
if label:
label_timestamps = label.timestamp
labels = label.label
closer_labels = []
closer_timestamps = []
for i in range(0, len(label_timestamps)):
min_dist = label_timestamps[len(label_timestamps) - 1] - label_timestamps[0]
closer_label = None
closer_timestamp = None
for heuristic_timestamp in timestamps:
dist = sqrt(np.square(label_timestamps[i] - heuristic_timestamp))
if dist < min_dist:
min_dist = dist
closer_label = labels[i]
closer_timestamp = heuristic_timestamp
closer_labels.append(closer_label)
closer_timestamps.append(closer_timestamp)
self.segment(sensors, closer_timestamps, closer_labels)
else:
self.segment(sensors, timestamps)
def get_statistical_features(self, data):
p = PeakAnalysis(self.view)
min_value = np.amin(data)
max_value = np.amax(data)
root_mean_square = UMath.get_root_mean_square(data)
peaks_number = np.mean(len(p.get_peaks(data)))
crest_factor = np.median(p.get_peak_to_average_ratios(data))
skewness = stats.skew(data, False)
kurtosis = stats.kurtosis(data, False)
variance = np.var(data)
return [min_value, max_value, root_mean_square, peaks_number, crest_factor, skewness, kurtosis, variance]
def analyzePlot(self,plotindex=-1): x,y=self.plot.plot.lines[plotindex].get_data() pylab.ion() fh=pylab.figure() pylab.plot(x,y,'ko-') CEN,FWHM,PEAK =PeakAnalysis.PeakAnalysis(x,y,nb=3,plotpoints=True) print 'Center=%e; FWHM=%e, PEAK=%e' %(CEN,FWHM,PEAK)
def fitgauss(x, y, e, a_init=None, x0_init=None, sigma_init=None, c_init=None):
n_bkg = 3
g = globals()
g["datax"] = x
g["datay"] = y
g["datae"] = e
(x0_guess, fwhm_guess, peak_guess) = PeakAnalysis.PeakAnalysis(x,
y,
nb=n_bkg)
if (x0_init is None): x0_init = x0_guess
if (sigma_init is None): sigma_init = fwhm_guess / 2.35
if (c_init is None):
c_init = (y[0:n_bkg].mean() + y[-1 - n_bkg:-1].mean()) / 2.
if (a_init is None): a_init = (y - c_init).max()
print x0_init, sigma_init, c_init, a_init
m = minuit2.Minuit2(chi2gauss,
a=a_init,
x0=x0_init,
sigma=sigma_init,
c=c_init)
m.printMode = 1
m.migrad()
fit_par = m.values
fit_err = m.errors
fit = fit_par["a"] * gauss(x, fit_par["x0"],
fit_par["sigma"]) + fit_par["c"]
return (fit_par, fit_err, x, fit)
def get_statistical_features(self, data):
p = PeakAnalysis(self.view)
min_value = np.amin(data)
max_value = np.amax(data)
root_mean_square = UMath.get_root_mean_square(data)
peaks_number = np.mean(len(p.get_peaks(data)))
crest_factor = np.median(p.get_peak_to_average_ratios(data))
skewness = stats.skew(data, False)
kurtosis = stats.kurtosis(data, False)
variance = np.var(data)
return [
min_value, max_value, root_mean_square, peaks_number, crest_factor,
skewness, kurtosis, variance
]
if pos == smaller:
new_data.append(data[smaller])
else:
# smaller == pos 인 경우 smaller == larger 이니까 이렇게 else를 larger = smaller + 1 으로
larger = smaller + 1
slope = data[larger] - data[smaller]
new_data.append(data[smaller] + slope * (pos - smaller))
new_data.append(data[peaks[-1]])
return new_data
if __name__ == '__main__':
inflation, unemployment = load_data()
# why not use mark real peaks, remove ambiguous peaks?
mod_inf, peak_inf = PeakAnalysis.find_peak(inflation, 0.6)
mod_une, peak_une = PeakAnalysis.find_peak(unemployment, 0.7)
matched = match_peaks(mod_inf, peak_inf, mod_une, peak_une)
interpolated_inf = interpolate_peaks(mod_inf,
list(map(lambda t: t[0], matched)))
interpolated_une = interpolate_peaks(mod_une,
list(map(lambda t: t[1], matched)))
assert (len(interpolated_inf) == len(interpolated_une))
# with EXPRESSION as NAME:
with open('interpolated_data.csv', 'w') as f:
for idx in range(len(interpolated_inf)):
f.write(
