def extractFeaturesWavelet(data, scales=[2, 4, 8, 16, 32], Class=0):
features = []
nObs, nSamp, nCols = data.shape
oClass = np.ones((nObs, 1)) * Class
for i in range(nObs):
scalo_features = np.array([])
for c in range(nCols):
scalo, scales = scalogram.scalogramCWT(data[i, :, c], scales)
scalo_features = np.append(scalo_features, scalo)
features.append(scalo_features)
return (np.array(features), oClass)
def extract_features_wavelet(data, empty_windows, scales=[2, 4, 8, 16, 32]):
features = []
n_obs_windows, n_samples, n_cols = data.shape
for i in range(n_obs_windows):
if i in empty_windows:
continue
scalogram_features = np.array([])
for c in range(n_cols):
scalo, fscales = scalogram.scalogramCWT(data[i, :, c], scales)
scalogram_features = np.append(scalogram_features, scalo)
features.append(scalogram_features)
return np.array(features)
waitforEnter()
# -13- #
import scalogram
for a in range(0, 40, 1):
plt.clf()
plt.figure(50)
x = data1[:, a]
scales = np.arange(1, 50)
plt.ion()
cwt = scalogram.CWTfft(x, scales)
plt.imshow(abs(cwt), cmap=plt.cm.Blues, aspect='auto')
plt.show()
plt.savefig("../imagens/scalogram/scalogramfft" + str(a) + ".png")
plt.clf()
plt.figure(51)
S, scales = scalogram.scalogramCWT(x, scales)
plt.plot(scales, S)
plt.show()
plt.savefig("../imagens/scalogram/scalogramCWT" + str(a) + ".png")
waitforEnter()
# End
waitforEnter()
f, psd = signal.periodogram(x) plt.plot(1 / f[:50], psd[:50]) plt.show() # -13- # import scalogram x = data1[:, 2] scales = np.arange(1, 50) plt.ion() plt.figure(11) cwt = scalogram.CWTfft(x, scales) plt.imshow(abs(cwt), cmap=plt.cm.Blues, aspect='auto') plt.show() plt.figure(12) S, scales = scalogram.scalogramCWT(x, scales) plt.plot(scales, S) plt.show() # -14- # # features M1 = np.mean(data1, axis=0) Md1 = np.median(data1, axis=0) V1 = np.var(data1, axis=0) S1 = stats.skew(data1) K1 = stats.kurtosis(data1) p = [25, 50, 75, 90, 95] Pr1 = np.array(np.percentile(data1, p, axis=0)).T M2 = np.mean(data2, axis=0) Md2 = np.median(data2, axis=0) V2 = np.var(data2, axis=0)