def test_upcoef_reconstruct():
data = np.arange(3)
a = pywt.downcoef("a", data, "haar")
d = pywt.downcoef("d", data, "haar")
rec = pywt.upcoef("a", a, "haar", take=3) + pywt.upcoef("d", d, "haar", take=3)
assert_allclose(rec, data)
def wavelet_dec(self, wavelet='db1', analyze_level=3):
sound = self.org_sound
coeffs = pywt.wavedec(sound, wavelet, level=analyze_level)
coef_a = coeffs[0]
coef_d = coeffs[1:]
approx_array = pywt.upcoef('a', coef_a, wavelet, level=analyze_level)
approx_array = [approx_array]
detail_array = list()
for ii in range(analyze_level):
detail_array.append(
pywt.upcoef('d',
coef_d[ii],
wavelet,
level=(analyze_level - ii)))
self.approx_array = approx_array
self.detail_array = detail_array
self.wave_mode = 'wavelet_dec'
self.level = analyze_level
approx = ['a' + str(self.level)]
detail = ['d' + str(self.level - ii) \
for ii in np.arange(self.level)]
col_name = approx + detail
self.col_name = col_name
return approx_array, detail_array
def wavelet_tree(self, wavelet='db1', analyze_level=6):
sound = self.org_sound
# 进行小波包分解
wp = pywt.WaveletPacket(data=sound, wavelet=wavelet, mode='sym')
# 获取小波包分解树的各个节点名称
node_names = [
node.path for node in wp.get_level(analyze_level, 'natural')
]
# 获取 概要部分节点名称 和 细节部分节点名称
node_names_approx = node_names[:(2**analyze_level // 2)]
node_names_detail = node_names[(2**analyze_level // 2):]
# 分别获取 概要部分的信号 和 细节部分的信号
approx_array = [pywt.upcoef('a', wp[node].data, wavelet, level=analyze_level) \
for node in node_names_approx]
detail_array = [pywt.upcoef('d', wp[node].data, wavelet, level=analyze_level) \
for node in node_names_detail]
self.approx_array = approx_array
self.detail_array = detail_array
self.wave_mode = 'wavelet_tree'
self.level = analyze_level
self.col_name = [node.path for node in \
wp.get_level(analyze_level, 'natural')]
return approx_array, detail_array
def py_wt(str, x, y):
signal = np.array(df.iloc[:, x:y]).reshape(1, -1)
signal_wt = pywt.wavedec(signal, 'db5', level=3)
signal_cA3, signal_cD3, signal_cD2, signal_cD1 = signal_wt
signal_cA3 = np.array(signal_cA3).flatten()
signal_cD3 = np.array(signal_cD3).flatten()
signal_cD2 = np.array(signal_cD2).flatten()
signal_cD1 = np.array(signal_cD1).flatten()
signal_cA3_ = pywt.upcoef('a', signal_cA3, 'db5', level=3, take=311)
signal_cD1_ = pywt.upcoef('d', signal_cD1, 'db5', level=1, take=311)
signal_cD2_ = pywt.upcoef('d', signal_cD2, 'db5', level=2, take=311)
signal_cD3_ = pywt.upcoef('d', signal_cD3, 'db5', level=3, take=311)
signal_cA3_ = signal_cA3_.tolist()
signal_cD1_ = signal_cD1_.tolist()
signal_cD2_ = signal_cD2_.tolist()
signal_cD3_ = signal_cD3_.tolist()
print("signal_cA3:")
print(signal_cA3)
df_signal = pd.DataFrame({
str + "_cA3": list(signal_cA3_),
str + "_cD3": list(signal_cD3_),
str + "_cD2": list(signal_cD2_),
str + "_cD1": list(signal_cD1_),
})
print("df_signal")
print(df_signal)
return df_signal
def test_upcoef_reconstruct():
data = np.arange(3)
a = pywt.downcoef('a', data, 'haar')
d = pywt.downcoef('d', data, 'haar')
rec = (pywt.upcoef('a', a, 'haar', take=3) +
pywt.upcoef('d', d, 'haar', take=3))
assert_allclose(rec, data)
def test_upcoef_complex():
rstate = np.random.RandomState(1234)
r = rstate.randn(4) + 1j * rstate.randn(4)
nlevels = 3
a = pywt.upcoef('a', r, 'haar', level=nlevels)
a_ref = pywt.upcoef('a', r.real, 'haar', level=nlevels)
a_ref = a_ref + 1j * pywt.upcoef('a', r.imag, 'haar', level=nlevels)
assert_allclose(a, a_ref)
def test_upcoef_complex():
rstate = np.random.RandomState(1234)
r = rstate.randn(4) + 1j*rstate.randn(4)
nlevels = 3
a = pywt.upcoef('a', r, 'haar', level=nlevels)
a_ref = pywt.upcoef('a', r.real, 'haar', level=nlevels)
a_ref = a_ref + 1j*pywt.upcoef('a', r.imag, 'haar', level=nlevels)
assert_allclose(a, a_ref)
def test_upcoef_reconstruct():
data = np.arange(3)
a = pywt.downcoef('a', data, 'haar')
d = pywt.downcoef('d', data, 'haar')
rec = (pywt.upcoef('a', a, 'haar', take=3) +
pywt.upcoef('d', d, 'haar', take=3))
assert_allclose(rec, data)
def test_upcoef_multilevel():
r = np.random.randn(4)
nlevels = 3
# calling with level=1 nlevels times
a1 = r.copy()
for i in range(nlevels):
a1 = pywt.upcoef('a', a1, 'haar', level=1)
# call with level=nlevels once
a3 = pywt.upcoef('a', r, 'haar', level=3)
assert_allclose(a1, a3)
def test_upcoef_multilevel():
r = np.random.randn(4)
nlevels = 3
# calling with level=1 nlevels times
a1 = r.copy()
for i in range(nlevels):
a1 = pywt.upcoef('a', a1, 'haar', level=1)
# call with level=nlevels once
a3 = pywt.upcoef('a', r, 'haar', level=3)
assert_allclose(a1, a3)
def wrcoef(X, coef_type, coeffs, wavename, level):
N = np.array(X).size
a, ds = coeffs[0], list(reversed(coeffs[1:]))
if coef_type =='a':
return pywt.upcoef('a', a, wavename, level=level)[:N]
elif coef_type == 'd':
return pywt.upcoef('d', ds[level-1], wavename, level=level)[:N]
else:
raise ValueError("Invalid coefficient type: {}".format(coef_type))
def wavelet_decomposition(y, wavelet_mother="db8", level_wt=3):
w = pywt.Wavelet(wavelet_mother)
coeffs = {}
wd = {}
coeffs['cA' + str(level_wt)] = pywt.downcoef('a', y, w, mode='sym', level=level_wt)
wd['cA' + str(level_wt)] = pywt.upcoef('a', coeffs['cA' + str(level_wt)], w, level=level_wt, take=len(y))
for i in range(level_wt, 0, -1):
coeffs['cD' + str(i)] = pywt.downcoef('d', y, w, mode='sym', level=i)
wd['cD' + str(i)] = pywt.upcoef('d', coeffs['cD' + str(i)], w, level=i, take=len(y))
return (w ,wd, coeffs)
def test_upcoef_docstring():
data = [1, 2, 3, 4, 5, 6]
(cA, cD) = pywt.dwt(data, 'db2', 'smooth')
rec = pywt.upcoef('a', cA, 'db2') + pywt.upcoef('d', cD, 'db2')
expect = [-0.25, -0.4330127, 1., 2., 3., 4., 5.,
6., 1.78589838, -1.03108891]
assert_allclose(rec, expect)
n = len(data)
rec = (pywt.upcoef('a', cA, 'db2', take=n) +
pywt.upcoef('d', cD, 'db2', take=n))
assert_allclose(rec, data)
def test_upcoef_multilevel():
rstate = np.random.RandomState(1234)
r = rstate.randn(4)
nlevels = 3
# calling with level=1 nlevels times
a1 = r.copy()
for i in range(nlevels):
a1 = pywt.upcoef("a", a1, "haar", level=1)
# call with level=nlevels once
a3 = pywt.upcoef("a", r, "haar", level=nlevels)
assert_allclose(a1, a3)
def test_upcoef_docstring():
data = [1, 2, 3, 4, 5, 6]
(cA, cD) = pywt.dwt(data, 'db2', 'smooth')
rec = pywt.upcoef('a', cA, 'db2') + pywt.upcoef('d', cD, 'db2')
expect = [-0.25, -0.4330127, 1., 2., 3., 4., 5.,
6., 1.78589838, -1.03108891]
assert_allclose(rec, expect)
n = len(data)
rec = (pywt.upcoef('a', cA, 'db2', take=n) +
pywt.upcoef('d', cD, 'db2', take=n))
assert_allclose(rec, data)
def feature_extraction(self, blocks):
print 'Feature extraction'
result = {"features": [], "tags": []}
i = 0
n = len(blocks)
level = 4
wavelet = 'db4'
for block in blocks:
feature = []
electrodes = [block.f3, block.f4, block.af3, block.af4]
for electrode in electrodes:
#Se obtienen las caracteristicas en el dominio del tiempo, media y desviacion estandar.
feature.append(np.mean(electrode))
feature.append(np.std(electrode))
size = len(electrode)
#Se aplica la transformada wavelet, para la separacion de las bandas de frecuencia.
wavelet_result = pywt.wavedec(electrode,
wavelet='db4',
level=level)
alpha = pywt.upcoef('d',
wavelet_result[2],
wavelet='db4',
level=3,
take=size)
beta = pywt.upcoef('d',
wavelet_result[3],
wavelet='db4',
level=2,
take=size)
gamma = pywt.upcoef('d',
wavelet_result[4],
wavelet='db4',
level=1,
take=size)
#Se obtienen las caracteristicas en el dominio de la frecuencia, media y desviacion estandar.
#ALPHA
feature.append(np.mean(self.transf(alpha)))
feature.append(np.std(self.transf(alpha)))
#BETA
feature.append(np.mean(self.transf(beta)))
feature.append(np.std(self.transf(beta)))
#GAMMA
feature.append(np.mean(self.transf(gamma)))
feature.append(np.std(self.transf(gamma)))
result['features'].append(feature)
result['tags'].append(block.tag)
return result
def __dwt(X):
m, n = X.shape
for i, x in enumerate(X):
ca, cd = pywt.dwt(x, 'db1')
tmp_a = pywt.upcoef('a', ca, 'db1', 1, n)
X[i, :] = tmp_a
return X
def training(self, file_data="user"):
raw = RawData(file_data)
blocks = raw.get_blocks_data()
block = self.get_block(blocks, 'SAD')
block_happy = self.get_block(blocks, 'HAPPY')
block_neutral = self.get_block(blocks, 'NEUTRAL')
#Definicion de los parametros para la aplicacion de la tranformada wavelet
level = 4
wavelet = 'db4'
sensors = [raw.f3, raw.f4, raw.af3, raw.af4]
for i in range(4):
n = len(sensors[i])
#Aplicacion de la transformada wavelet.
cA4, cD4, cD3, cD2, cD1 = pywt.wavedec(sensors[i],
wavelet=wavelet,
level=level)
#Extraccion de las 5 bandas de frecuencia.
delta = pywt.upcoef('a', cA4, wavelet=wavelet, level=level, take=n)
theta = pywt.upcoef('d', cD4, wavelet=wavelet, level=level, take=n)
alpha = pywt.upcoef('d', cD3, wavelet=wavelet, level=3, take=n)
beta = pywt.upcoef('d', cD2, wavelet=wavelet, level=2, take=n)
gamma = pywt.upcoef('d', cD1, wavelet=wavelet, level=1, take=n)
#Despligue de la grafica luego de la aplicacion de la transformada
plt.figure(i, figsize=(20, 10))
plt.subplot(511)
plt.plot(self.transf(delta))
plt.ylabel("delta")
plt.subplot(512)
plt.plot(self.transf(theta))
plt.ylabel("theta")
plt.subplot(513)
plt.plot(self.transf(alpha))
plt.ylabel("alpha")
plt.subplot(514)
plt.plot(self.transf(beta))
plt.ylabel("beta")
plt.subplot(515)
plt.plot(self.transf(gamma))
plt.ylabel("gamma")
plt.savefig('../../imagenes/F' + str(i + 1) + 'ANEXOS_' + user +
'.png')
plt.show()
def plotWavelet(self, wavelet, level):
coeffs = pywt.wavedec(self.y, wavelet, level=level)
self.yWavelet = pywt.upcoef('a',
coeffs[0],
wavelet,
level=level,
take=len(self.y))
name = '{}, level:{}'.format(wavelet, str(level))
plt.plot(self.x, self.yWavelet, label=name)
def mydwt(y):
"""離散ウェーブレット分解
"""
ca, cd = pywt.dwt(y.values, 'db1')
n = len(y)
columns = ['dwt_approx', 'dwt_detail']
y_dwt = pd.DataFrame(np.zeros([n,len(columns)]), columns=columns, index=y.index)
y_dwt['dwt_approx'] = pywt.upcoef('a', ca, 'db1')
y_dwt['dwt_detail'] = pywt.upcoef('d', cd, 'db1')
#y_dwt.columns = [columns[0] + '_' + columns[1] for i in range(len(columns))]
#feature_labels = [[y.name, columns[i]] for i in range(len(columns))]
feature_labels = [[str(y.name), columns[i]] for i in range(len(columns))]
y_dwt.columns = [str(y.name) + '_' + columns[i] for i in range(len(columns))]
return y_dwt, feature_labels
def reconstruct_without_approx(xs, labels, level, fig=None):
# reconstruct
rs = [pywt.upcoef('d', x, 'db4', level=level) for x in xs]
# generate plot
if fig is None:
fig = plt.figure()
for i, x in enumerate(xs):
plt.plot((np.abs(x))**2, label="Power of reconstructed signal ({})".format(labels[i]))
plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)
return rs, fig
def py_wt(str, x, y):
signal = np.array(df.iloc[:, x:y]).reshape(1, -1)
signal_wt = pywt.wavedec(signal, 'db5', level=3)
signal_cA3, signal_cD3, signal_cD2, signal_cD1 = signal_wt
signal_cA3 = np.array(signal_cA3).flatten()
signal_cD3 = np.array(signal_cD3).flatten()
signal_cD2 = np.array(signal_cD2).flatten()
signal_cD1 = np.array(signal_cD1).flatten()
signal_cA3_ = pywt.upcoef('a', signal_cA3, 'db5', level=3, take=311)
signal_cD1_ = pywt.upcoef('d', signal_cD1, 'db5', level=1, take=311)
signal_cD2_ = pywt.upcoef('d', signal_cD2, 'db5', level=2, take=311)
signal_cD3_ = pywt.upcoef('d', signal_cD3, 'db5', level=3, take=311)
signal_cA3_ = signal_cA3_.tolist()
signal_cD1_ = signal_cD1_.tolist()
signal_cD2_ = signal_cD2_.tolist()
signal_cD3_ = signal_cD3_.tolist()
return signal_cA3_,signal_cD3_,signal_cD2_,signal_cD1_
def next_lvl(self) -> None:
i = self.start_lvl - self.curr_lvl
self.cur_cA = self.cur_cA + self.cD[i]
self.cur_cA = pywt.upcoef('a',
self.cur_cA,
self.filter_type,
take=len(self.cD[i + 1]))
self.curr_lvl -= 1
self.recunstruct = 2**self.curr_lvl
self.k *= 2
return
def wavelet_transform(wave, wavelet, level):
# Wavelet Transform
wavelet = pywt.Wavelet(wavelet)
coeffs = pywt.wavedec(wave, wavelet=wavelet, level=level)
details_reconstructed = np.zeros(len(wave))
for i in range(level):
details_reconstructed = np.vstack((details_reconstructed,
pywt.upcoef(part='d',
coeffs=coeffs[i + 1],
wavelet=wavelet,
level=level - (i),
take=len(wave))))
return details_reconstructed[1:]
def wavelet_packet_rec(X, wavelet, max_level=3):
"""Return the node wise reconstructions of a full wavelet decompositions.
:param X: Array of size (n, d), where `d` is the number of different signals;
each will be decomposed separately.
:param wavelet: Wavelet to use; string will be passed on to PyWavelets.
:param max_level: Maximum depth of the resulting packet tree.
:returns: Array of size `(n, d * 2**3)`.
"""
res = []
for i in range(X.shape[1]):
x = X[:, i]
wp = pywt.WaveletPacket(x, wavelet, maxlevel=max_level)
for node in wp.get_leaf_nodes(True):
res.append(pywt.upcoef(node.path[-1], node.data, wavelet, node.level))
return np.concatenate([i[:, np.newaxis] for i in res], axis=1)
def get_feature_2(data):
# fluc, seasonal_c, local_c
logging.info('get feature v2')
feature = np.empty((data.shape[0], 3 * config.x_dim), dtype=np.float)
for i in range(config.x_dim):
col = data[:, i]
fluc = col.std() / col.mean() if col.mean() > 0 else col.std()
diff = col[1440:] - col[:-1440]
seasonal_c = diff.std()
cA1, cD1 = pywt.dwt(col, 'db2')
D1 = pywt.upcoef('d', cD1, 'db2', 1)
local_c = D1.std()
feature[:, i * 3] = [fluc] * len(data)
feature[:, i * 3 + 1] = [seasonal_c] * len(data)
feature[:, i * 3 + 2] = [local_c] * len(data)
return feature
def ReconUsingUpcoef(signal, wname='db4', level=3):
w = Wavelet(wname)
X = signal
#coeffs = wavedec(a, w, mode='symmetric', level=level, axis=-1)
coeffs = []
for i in range(level):
a, d = dwt(X, w, mode='symmetric', axis=-1)
coeffs.append(d)
coeffs.append(a)
coeffs.reverse()
ca = coeffs[0]
n = len(X)
return upcoef('a', ca, w, take=n, level=level)
def DwtRec(signal, wavelet, level):
# 返回的是每一层的分量，即将原始信号进行了频带分解，时间长度与原始序列signal等长
# 每一层对应一行
CoeffsRev = []
Coeffs = DWT(signal, wavelet=wavelet, level=level)
for row in range(level + 1):
if row == 0:
# 重构近似系数，即低频分量
leveltype = level
part = 'a'
else:
# 重构细节系数，即高频分量
leveltype = level - row + 1
part = 'd'
EachLevelRev = pywt.upcoef(part,
Coeffs[row],
wavelet=wavelet,
level=leveltype,
take=len(signal))
CoeffsRev.append(EachLevelRev)
return np.array(CoeffsRev)
def wv_signal_decomprs(self, signal_df):
"""
Make a compression of the input signals with a discrete wavelet
transform.
"""
signal_wv_list = []
# and decompress data effectively. take=self.original_signal_len
for signal_name in self.signals_names:
temp_wv_list = pywt.upcoef('a',
signal_df[signal_name],
self.wv_sel,
level=self.wv_order).tolist()
signal_wv_list.append(temp_wv_list)
print("Signal length after decompression: {}".format(
len(signal_wv_list[0])))
# Create and return the output dataframe with the compressed signals
output_df = pd.DataFrame()
for i, signal in enumerate(self.signals_names):
output_df[signal] = \
self.correct_deco_len(self.original_signal_df[signal],
signal_wv_list[i])
return output_df
# For the scaling function, we need to compute the inverse DWT with a delta
# for the approximation coefficients. (All detail coefficients are set
# to zero.)
# EY : 20150703 note on using PyWavelets
# to make a custom Wavelet, you need to specify lowpass and highpass decomposition filters and lowpass and highpass reconstruction filters for a Filter Bank
# how do we get the lowpass and highpass decomposition filters from the highpass reconstruction filters?
# see my note in my pdf/LaTeX writeup, note on Sec. 6.2. of Strang and Nguyen
# the answer is to just reverse the order cf. http://www.pybytes.com/pywavelets/regression/wavelet.html
c = f0[::-1]
d = f1[::-1]
dec_lo, dec_hi, rec_lo, rec_hi = np.hstack(c), np.hstack(d), np.hstack(f0), np.hstack(f1)
filter_bank = [dec_lo, dec_hi, rec_lo, rec_hi ]
DaubechiesWavelet = pywt.Wavelet(name="DaubechiesWavelet", filter_bank=filter_bank)
y = pywt.upcoef('a', [1.,0.], DaubechiesWavelet, numlevels)
phi = M*np.append(0, y[0:L])
# For the wavelet, we need to compute the inverse DWT with a delta for the
# detail coefficients. (All approximation coefficients and all detail
# coefficients at finer scales are set to zero.)
y = pywt.upcoef('d', [1.,0.], DaubechiesWavelet, numlevels) # Inverse DWT
w = M*np.append(0,y[0:L])
# Determine the time vector.
t = np.vstack(np.array(range(L+1)))/M
# Plot the results.
plt.figure(1)
plt.plot(t,phi,'-',label="Scaling function")
plt.plot(t,w,'--',label="Wavelet")
import pywt
#--信号data---
cA, cD1, cD2, cD3 = pywt.wavedec(data, 'haar', level=3) #分解
data = pywt.waverec([cA, cD1, cD2, cD3], 'haar') #重构
x1 = pywt.upcoef('a', cA, 'haar', level=3, take=N) #计算第3层尺度系数的信号，长度为N
x2 = pywt.upcoef('d', cD1, 'haar', level=3, take=N) #计算第3层细节系数的信号，长度为N
def Beat_analysis_preproc(*args):
nargin = len(args)
if nargin == 0:
print('pas d' 'entree correcte pour la location de l' 'onde R')
return
elif nargin == 1:
signal = args[0]
level_max_BLD = 8
wavelet_BLD = 'sym4'
build_level_BLD = 8
level_max_DEN = 3
wavelet_DEN = 'sym4'
elif nargin == 2:
signal = args[0]
level_max_BLD = args[1]
wavelet_BLD = 'sym4'
build_level_BLD = args[1]
level_max_DEN = 3
wavelet_DEN = 'sym4'
elif nargin == 3:
signal = args[0]
level_max_BLD = args[1]
wavelet_BLD = args[2]
build_level_BLD = args[1]
level_max_DEN = 3
wavelet_DEN = 'sym4'
elif nargin == 4:
signal = args[0]
level_max_BLD = args[1]
wavelet_BLD = args[2]
build_level_BLD = args[3]
level_max_DEN = 3
wavelet_DEN = 'sym4'
elif nargin == 5:
signal = args[0]
level_max_BLD = args[1]
wavelet_BLD = args[2]
build_level_BLD = args[3]
level_max_DEN = args[4]
wavelet_DEN = 'sym4'
else:
signal = args[0]
level_max_BLD = args[1]
wavelet_BLD = args[2]
build_level_BLD = args[3]
level_max_DEN = args[4]
wavelet_DEN = args[5]
varargout = list()
if level_max_BLD:
coeffs = pywt.wavedec(signal, wavelet_BLD, 'sym', level_max_BLD)
baseline = pywt.upcoef('a', coeffs[0], 'sym4', 8,
take=len(signal) + 6)[:int(len(signal))]
signal_basrem = signal - baseline
varargout.append(signal_basrem)
varargout.append(baseline)
else:
signal_basrem = signal
varargout.append(signal_basrem)
varargout.append(np.zeros((len(signal))))
if level_max_DEN:
deb = signal[0]
sig_basrem_d1 = WDen.wden(signal_basrem - deb, 'sqtwolog', 'soft',
'sln', level_max_DEN, wavelet_DEN) + deb
sig_basrem_d2 = WDen.wden(signal_basrem - deb, 'minimaxi', 'soft',
'sln', level_max_DEN, wavelet_DEN) + deb
varargout.append(sig_basrem_d1)
varargout.append(sig_basrem_d2)
else:
varargout.append(signal_basrem)
varargout.append(signal_basrem)
return varargout[0], varargout[1], varargout[2], varargout[3]
df_day_of_week = df.iloc[:, 7:8]
df_holiday = df.iloc[:, 8:9]
df_load = df.iloc[:, 9:10]
printc("用db5小波进行yes_ave 3层分解")
yes_ave = np.array(df.iloc[:, 1:2]).reshape(1, -1)
# print(ML.shape)
yes_ave_wt = pywt.wavedec(yes_ave, 'db5', level=3)
yes_ave_cA3, yes_ave_cD3, yes_ave_cD2, yes_ave_cD1 = yes_ave_wt
yes_ave_cA3 = np.array(yes_ave_cA3).flatten()
yes_ave_cD3 = np.array(yes_ave_cD3).flatten()
yes_ave_cD2 = np.array(yes_ave_cD2).flatten()
yes_ave_cD1 = np.array(yes_ave_cD1).flatten()
yes_ave_cA3_ = pywt.upcoef('a', yes_ave_cA3, 'db5', level=3, take=311)
yes_ave_cD1_ = pywt.upcoef('d', yes_ave_cD1, 'db5', level=1, take=311)
yes_ave_cD2_ = pywt.upcoef('d', yes_ave_cD2, 'db5', level=2, take=311)
yes_ave_cD3_ = pywt.upcoef('d', yes_ave_cD3, 'db5', level=3, take=311)
yes_ave_cA3_ = yes_ave_cA3_.tolist()
yes_ave_cD1_ = yes_ave_cD1_.tolist()
yes_ave_cD2_ = yes_ave_cD2_.tolist()
yes_ave_cD3_ = yes_ave_cD3_.tolist()
print(yes_ave_cA3)
printc("用db5小波进行yes_load 3层分解")
yes_load = np.array(df.iloc[:, 2:3]).reshape(1, -1)
# print(ML.shape)
yes_load_wt = pywt.wavedec(yes_load, 'db5', level=3)
# to zero.)
# EY : 20150703 note on using PyWavelets
# to make a custom Wavelet, you need to specify lowpass and highpass decomposition filters and lowpass and highpass reconstruction filters for a Filter Bank
# how do we get the lowpass and highpass decomposition filters from the highpass reconstruction filters?
# see my note in my pdf/LaTeX writeup, note on Sec. 6.2. of Strang and Nguyen
# the answer is to just reverse the order cf. http://www.pybytes.com/pywavelets/regression/wavelet.html
c = f0[::-1]
d = f1[::-1]
dec_lo, dec_hi, rec_lo, rec_hi = np.hstack(c), np.hstack(d), np.hstack(
f0), np.hstack(f1)
filter_bank = [dec_lo, dec_hi, rec_lo, rec_hi]
DaubechiesWavelet = pywt.Wavelet(name="DaubechiesWavelet",
filter_bank=filter_bank)
y = pywt.upcoef('a', [1., 0.], DaubechiesWavelet, numlevels)
phi = M * np.append(0, y[0:L])
# For the wavelet, we need to compute the inverse DWT with a delta for the
# detail coefficients. (All approximation coefficients and all detail
# coefficients at finer scales are set to zero.)
y = pywt.upcoef('d', [1., 0.], DaubechiesWavelet, numlevels) # Inverse DWT
w = M * np.append(0, y[0:L])
# Determine the time vector.
t = np.vstack(np.array(range(L + 1))) / M
# Plot the results.
plt.figure(1)
plt.plot(t, phi, '-', label="Scaling function")
plt.plot(t, w, '--', label="Wavelet")
# df.to_excel("cjp2.xlsx",sheet_name="xcjp",index=False,header=True)
##########################################################################
##########################################################################
printc("用db5小波进行日前发电量3层分解")
DAG = np.array(df.iloc[:, 5:6]).reshape(1, -1)
# print(ML.shape)
DAG_wt = pywt.wavedec(DAG, 'db5', level=3)
DAG_cA3, DAG_cD3, DAG_cD2, DAG_cD1 = DAG_wt
DAG_cA3 = np.array(DAG_cA3).flatten()
DAG_cD3 = np.array(DAG_cD3).flatten()
DAG_cD2 = np.array(DAG_cD2).flatten()
DAG_cD1 = np.array(DAG_cD1).flatten()
DAG_cA3_ = pywt.upcoef('a', DAG_cA3, 'db5', level=3, take=311)
DAG_cD1_ = pywt.upcoef('d', DAG_cD1, 'db5', level=1, take=311)
DAG_cD2_ = pywt.upcoef('d', DAG_cD2, 'db5', level=2, take=311)
DAG_cD3_ = pywt.upcoef('d', DAG_cD3, 'db5', level=3, take=311)
DAG_cA3_ = DAG_cA3_.tolist()
DAG_cD1_ = DAG_cD1_.tolist()
DAG_cD2_ = DAG_cD2_.tolist()
DAG_cD3_ = DAG_cD3_.tolist()
printc("用db5小波进行周前发电量3层分解")
WC = np.array(df.iloc[:, 4:5]).reshape(1, -1)
WC_wt = pywt.wavedec(WC, 'db5', level=3)
WC_cA3, WC_cD3, WC_cD2, WC_cD1 = WC_wt
WC_cA3 = np.array(WC_cA3).flatten()
def wavelet_transform(data, wavelet='db2', mode='reflect'):
(cA, cD) = pywt.dwt(data, wavelet, mode)
return pywt.upcoef('a',cA,wavelet,take = len(data)), \
pywt.upcoef('d',cD,wavelet,take = len(data))
def progressive_lowcut_series(series):
"""Progressively remove low-frequency components of 1D series.
Yields sequence in which the low-frequency (large-scale) components
of `series` are progressively removed. The sequence is obtained
from reconstrution of a multilevel discrete haar wavelet
decompositon of `series`.
N.B.: The length of `series` is assumed to be a power of 2
(does not check!)
Parameters
----------
series : array_like
1D data series with a length that is a power of 2
Yields
-------
lowcut_series : array
Sequence of progressively lowcut filtered data `series`. The
yielded series have the same length as `series`.
Notes
-----
After an N level discrete wavelet decomposition, a data series S can
be reconstructed in terms of wavelet 'approximations' (A) and
'details' (D)::
S = A(N) + D(N) + D(N-1) ... D(2) + D(1)
= A(N-1) + D(N-1) + ... D(2) + D(1) [A(N-1) = A(N) + D(N)]
...
= A(1) + D(1) [(A(1) = A(2) + D(2)]
where A(N) represents the 'lowest' level approximation [e.g., for
the haar wavelet and a complete decomposition of dyadic length data,
A(N) is equal to mean(S)]
The sequence returned by this function is::
S - A(N),
S - A(N-1),
S - A(N-2),
...,
S - A(1)
This sequence is computed by the equivalent::
S - A(N),
S - A(N) - D(N),
S - A(N) - D(N) - D(N-1),
...,
S - A(N) - D(N) - D(N-1) - ... - D(2),
i.e. the details are removed in sequence::
lowcut = S - A(N)
for j = N to 2
lowcut = lowcut - D(j)
"""
series_data = np.asarray(series)
wavelet = pywt.Wavelet('haar')
nlevel = pywt.dwt_max_level(series_data.size, wavelet.dec_len)
decomp_coef = pywt.wavedec(series_data, wavelet=wavelet, level=nlevel)
cAn, cD = decomp_coef[0], decomp_coef[1:]
lowcut_series = series_data - pywt.upcoef('a', cAn, wavelet, level=nlevel)
yield lowcut_series
for j, cDj in enumerate(cD[:-1]):
Dj = pywt.upcoef('d', cDj, wavelet, level=nlevel - j)
lowcut_series -= Dj
yield lowcut_series
def approx(x, wavelet, level):
ca = pywt.wavedec(x, wavelet, level=level)
ca = ca[0]
return pywt.upcoef('a', ca, wavelet, level, take=len(x))
def progressive_lowcut_series(series):
"""Progressively remove low-frequency components of 1D series.
Yields sequence in which the low-frequency (large-scale) components
of `series` are progressively removed. The sequence is obtained
from reconstruction of a multilevel discrete haar wavelet
decomposition of `series`.
N.B.: The length of `series` is assumed to be a power of 2
(does not check!)
Parameters
----------
series : array_like
1D data series with a length that is a power of 2
Yields
-------
lowcut_series : array
Sequence of progressively lowcut filtered data `series`. The
yielded series have the same length as `series`.
Notes
-----
After an N level discrete wavelet decomposition, a data series S can
be reconstructed in terms of wavelet 'approximations' (A) and
'details' (D)::
S = A(N) + D(N) + D(N-1) ... D(2) + D(1)
= A(N-1) + D(N-1) + ... D(2) + D(1) [A(N-1) = A(N) + D(N)]
...
= A(1) + D(1) [(A(1) = A(2) + D(2)]
where A(N) represents the 'lowest' level approximation. For
the haar wavelet and data S having a length that is a power of 2,
A(N) is equal to mean(S)]
The sequence returned by this function is::
S - A(N),
S - A(N-1),
S - A(N-2),
...,
S - A(1)
This sequence is computed by the equivalent::
S - A(N),
S - A(N) - D(N),
S - A(N) - D(N) - D(N-1),
...,
S - A(N) - D(N) - D(N-1) - ... - D(2),
i.e. the details are removed in sequence::
lowcut = S - A(N)
for j = N to 2
lowcut -= D(j)
"""
series_data = np.asarray(series)
wavelet = pywt.Wavelet("haar")
nlevel = pywt.dwt_max_level(series_data.size, wavelet.dec_len)
decomp_coef = pywt.wavedec(series_data, wavelet=wavelet, level=nlevel)
cAn, cD = decomp_coef[0], decomp_coef[1:]
lowcut_series = series_data - pywt.upcoef("a", cAn, wavelet, level=nlevel)
yield lowcut_series
for j, cDj in enumerate(cD[:-1]):
Dj = pywt.upcoef("d", cDj, wavelet, level=nlevel - j)
lowcut_series -= Dj
yield lowcut_series
