kopia lustrzana https://github.com/gabrielegilardi/SignalFilters
add: kalman filter, ABG filters, synthetic wave
Gabriele Gilardi
rodzic
8fef296eba
commit
d986665244
|
|
@ -1,183 +0,0 @@
|
|||
|
||||
# import numpy as np
|
||||
|
||||
# measurements = np.array([5., 6., 7., 9., 10.])
|
||||
# motion = np.array([1., 1., 2., 1., 1.])
|
||||
# measurement_sigma = 4.
|
||||
# motion_sigma = 2.
|
||||
# mu = 0.
|
||||
# sigma = 1000.
|
||||
|
||||
# # Measurement
|
||||
# def Update( mean1, var1, mean2, var2 ):
|
||||
# mean = (var2*mean1 + var1*mean2) / (var1 + var2)
|
||||
# var = 1.0 / (1.0/var1 + 1.0/var2)
|
||||
# return [mean, var]
|
||||
|
||||
# # Motion
|
||||
# def Predict( mean1, var1, U, varU ):
|
||||
# mean = mean1 + U
|
||||
# var = var1 + varU
|
||||
# return [mean, var]
|
||||
|
||||
# for n in range(len(measurements)):
|
||||
# [mu, sigma] = Update(mu, sigma, measurements[n], measurement_sigma)
|
||||
# print('Update : ', n, [mu, sigma])
|
||||
# [mu, sigma] = Predict(mu, sigma, motion[n],motion_sigma)
|
||||
# print('Predict: ', n, [mu, sigma])
|
||||
|
||||
# print(' ')
|
||||
# print(Update(1,1,3,1))
|
||||
|
||||
# -------------------------------------------------------
|
||||
|
||||
# import numpy as np
|
||||
|
||||
# measurements = [ 1., 2., 3. ]
|
||||
# dt = 1.
|
||||
|
||||
# # Initial state (location and velocity)
|
||||
# x = np.array([[ 0. ],
|
||||
# [ 0. ]])
|
||||
# # Initial uncertainty
|
||||
# P = np.array([[ 1000., 0. ],
|
||||
# [ 0., 1000. ]])
|
||||
# # External motion
|
||||
# U = np.array([[ 0. ],
|
||||
# [ 0. ]])
|
||||
# # Next state function
|
||||
# F = np.array([[ 1., dt ],
|
||||
# [ 0., 1. ]])
|
||||
# # Measurement function
|
||||
# H = np.array([[ 1., 0. ]])
|
||||
# # Measurement uncertainty
|
||||
# R = np.array([[ 1. ]])
|
||||
# # Identity matrix
|
||||
# I = np.eye(2)
|
||||
|
||||
|
||||
# def filter(x, P):
|
||||
|
||||
# step = 0
|
||||
# for z in (measurements):
|
||||
# step += 1
|
||||
# print("step = ", step, " meas. = ", z)
|
||||
|
||||
# # Measurement
|
||||
# Htra = H.T
|
||||
# S = H.dot(P.dot(Htra)) + R
|
||||
# Sinv = np.linalg.inv(S)
|
||||
# K = P.dot(Htra.dot(Sinv))
|
||||
# y = z - H.dot(x)
|
||||
# xp = x +K.dot(y)
|
||||
# Pp = P - K.dot(H.dot(P))
|
||||
|
||||
# # Prediction
|
||||
# x = F.dot(xp) + U
|
||||
# Ftra = F.T
|
||||
# P = F.dot(Pp.dot(Ftra))
|
||||
|
||||
# print('x =')
|
||||
# print(x)
|
||||
# print('P =')
|
||||
# print(P)
|
||||
|
||||
# filter(x, P)
|
||||
|
||||
# # -------------------------------------------------------
|
||||
|
||||
import numpy as np
|
||||
|
||||
x0 = 4.
|
||||
y0 = 12.
|
||||
measurements = np.array([[ 5., 10. ],
|
||||
[ 6., 8. ],
|
||||
[ 7., 6. ],
|
||||
[ 8., 4. ],
|
||||
[ 9., 2. ],
|
||||
[ 10., 0. ]])
|
||||
# x0 = -4.
|
||||
# y0 = 8.
|
||||
# measurements = np.array([[ 1., 4. ],
|
||||
# [ 6., 0. ],
|
||||
# [ 11., -4. ],
|
||||
# [ 16., -8. ]])
|
||||
# x0 = 1.
|
||||
# y0 = 19.
|
||||
# measurements = np.array([[ 1., 17. ],
|
||||
# [ 1., 15. ],
|
||||
# [ 1., 13. ],
|
||||
# [ 1., 11. ]])
|
||||
# x0 = 1.
|
||||
# y0 = 19.
|
||||
# measurements = np.array([[ 2., 17. ],
|
||||
# [ 0., 15. ],
|
||||
# [ 2., 13. ],
|
||||
# [ 0., 11. ]])
|
||||
# Time step
|
||||
dt = 0.1
|
||||
# Initial state (location and velocity)
|
||||
x = np.array([[ x0 ],
|
||||
[ y0 ],
|
||||
[ 0. ],
|
||||
[ 0. ]])
|
||||
# Initial uncertainty
|
||||
P = np.array([[ 0., 0., 0., 0. ],
|
||||
[ 0., 0., 0., 0. ],
|
||||
[ 0., 0., 1000., 0. ],
|
||||
[ 0., 0., 0., 1000. ]])
|
||||
# External motion
|
||||
U = np.array([[ 0. ],
|
||||
[ 0. ],
|
||||
[ 0. ],
|
||||
[ 0. ]])
|
||||
# Next state function
|
||||
F = np.array([[ 1., 0., dt, 0. ],
|
||||
[ 0., 1., 0., dt ],
|
||||
[ 0., 0., 1., 0. ],
|
||||
[ 0., 0., 0., 1. ]])
|
||||
# Measurement function
|
||||
H = np.array([[ 1., 0., 0., 0. ],
|
||||
[ 0., 1., 0., 0. ]])
|
||||
# Measurement uncertainty
|
||||
R = np.array([[ 0.1, 0. ],
|
||||
[ 0. , 0.1 ]])
|
||||
# Measurement vector
|
||||
z = np.zeros((2,1))
|
||||
|
||||
|
||||
def filter(x, P):
|
||||
|
||||
for n in range(len(measurements)):
|
||||
|
||||
z[0][0] = measurements[n][0]
|
||||
z[1][0] = measurements[n][1]
|
||||
|
||||
# Prediction
|
||||
xp = F.dot(x) + U
|
||||
Ftra = F.T
|
||||
Pp = F.dot(P.dot(Ftra))
|
||||
|
||||
# Measurement
|
||||
Htra = H.T
|
||||
S = H.dot(Pp.dot(Htra)) + R
|
||||
Sinv = np.linalg.inv(S)
|
||||
K = Pp.dot(Htra.dot(Sinv))
|
||||
y = z - H.dot(xp)
|
||||
x = xp +K.dot(y)
|
||||
P = Pp - K.dot(H.dot(Pp))
|
||||
# print(z)
|
||||
# print('x = ')
|
||||
# print(x)
|
||||
# print('P = ')
|
||||
# print(P)
|
||||
# print(' ')
|
||||
|
||||
return x, P
|
||||
|
||||
|
||||
x_final, P_final = filter(x, P)
|
||||
print('x = ')
|
||||
print(x_final)
|
||||
print('P = ')
|
||||
print(P_final)
|
||||
|
|
@ -1,106 +0,0 @@
|
|||
"""
|
||||
|
||||
correction = update = measurement
|
||||
prediction = motion
|
||||
|
||||
X (n_states, 1) State vector
|
||||
P (n_states, n_states) Covariance matrix of X
|
||||
F (n_states, n_states) State transition matrix
|
||||
U (n_states, 1) Input/control/drift vector
|
||||
Z (n_meas, 1) Measurament vector
|
||||
H (n_meas, n_states) Measurament matrix
|
||||
R (n_meas, n_meas) Covariance matrix of Z
|
||||
S (n_meas, n_meas) Covariance matrix (?)
|
||||
K (n_states, m) Kalman gain
|
||||
Q (n_states, n_states) Covariance matrix (?)
|
||||
|
||||
Data (n_meas, n_samples) Measurements
|
||||
Fext (n_states, n_samples) External driver
|
||||
X0 (n_states, 1) Initial state vector
|
||||
P0 (n_states, n_states) Initial covariance matrix of X
|
||||
"""
|
||||
|
||||
import numpy as np
|
||||
|
||||
|
||||
class KalmanFilter:
|
||||
|
||||
def __init__(self, F, H, Q, R):
|
||||
"""
|
||||
"""
|
||||
self.F = F
|
||||
self.Q = Q
|
||||
self.H = H
|
||||
self.R = R
|
||||
|
||||
|
||||
def prediction(self, X, P, U):
|
||||
X = self.F @ X + U
|
||||
P = self.F @ P @ self.F.T + self.Q
|
||||
return X, P
|
||||
|
||||
|
||||
def update(self, X, P, Z):
|
||||
"""
|
||||
"""
|
||||
S = self.H @ P @ self.H.T + self.R
|
||||
K = P @ self.H.T @ np.linalg.inv(S)
|
||||
X = X + K @ (Z - self.H @ X)
|
||||
P = P - K @ self.H @ P
|
||||
return X, P
|
||||
|
||||
|
||||
def applyFilter(self, Data, Fext, X0, P0):
|
||||
"""
|
||||
"""
|
||||
pass
|
||||
|
||||
|
||||
# Measurements
|
||||
data = np.array([[5., 6., 7., 8., 9., 10.],
|
||||
[10., 8., 6., 4., 2., 0.]])
|
||||
|
||||
# Initial state vector
|
||||
X0 = np.array([[4. ],
|
||||
[12.],
|
||||
[0. ],
|
||||
[0. ]])
|
||||
|
||||
# Initial covariance matrix of X
|
||||
P0 = np.array([[0., 0., 0., 0.],
|
||||
[0., 0., 0., 0.],
|
||||
[0., 0., 1000., 0.],
|
||||
[0., 0., 0., 1000.]])
|
||||
|
||||
# External motion
|
||||
Fext = np.zeros_like(data)
|
||||
|
||||
# Next state function
|
||||
dt = 0.1
|
||||
F = np.array([[ 1., 0., dt, 0. ],
|
||||
[ 0., 1., 0., dt ],
|
||||
[ 0., 0., 1., 0. ],
|
||||
[ 0., 0., 0., 1. ]])
|
||||
# Measurement function
|
||||
H = np.array([[ 1., 0., 0., 0. ],
|
||||
[ 0., 1., 0., 0. ]])
|
||||
# Measurement uncertainty
|
||||
R = np.array([[ 0.1, 0. ],
|
||||
[ 0. , 0.1 ]])
|
||||
|
||||
def filter(x, P):
|
||||
|
||||
step = 0
|
||||
for z in (measurements):
|
||||
step += 1
|
||||
print("step = ", step, " meas. = ", z)
|
||||
|
||||
# Update
|
||||
|
||||
|
||||
print('x =')
|
||||
print(x)
|
||||
print('P =')
|
||||
print(P)
|
||||
|
||||
filter(x, P)
|
||||
|
|
@ -46,6 +46,8 @@ InstTrend N/alpha Instantaneous trendline
|
|||
SincFunction N Sinc function
|
||||
Decycler P Decycler, 1-GaussHigh (P>=5)
|
||||
DecyclerOsc P1,P2 Decycle oscillator, GH(P1) - GH(P2), (P1>=5)
|
||||
ABG
|
||||
Kalman
|
||||
|
||||
N Order/smoothing factor/number of previous samples
|
||||
alpha Damping term
|
||||
|
|
@ -87,12 +89,12 @@ def filter_data(X, b, a):
|
|||
|
||||
class Filter:
|
||||
|
||||
def __init__(self, X):
|
||||
def __init__(self, data):
|
||||
"""
|
||||
"""
|
||||
self.X = np.asarray(X)
|
||||
self.data = np.asarray(data)
|
||||
|
||||
self.n_samples, self.n_series = X.shape
|
||||
self.n_samples, self.n_series = data.shape
|
||||
self.idx = 0
|
||||
|
||||
def Generic(self, b=1.0, a=1.0):
|
||||
|
|
@ -101,7 +103,7 @@ class Filter:
|
|||
"""
|
||||
b = np.asarray(b)
|
||||
a = np.asarray(a)
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def SMA(self, N=10):
|
||||
|
|
@ -110,7 +112,7 @@ class Filter:
|
|||
"""
|
||||
b = np.ones(N) / N
|
||||
a = np.array([1.0])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def EMA(self, N=10, alpha=None):
|
||||
|
|
@ -123,7 +125,7 @@ class Filter:
|
|||
alpha = 2.0 / (N + 1.0)
|
||||
b = np.array([alpha])
|
||||
a = np.array([1.0, -(1.0 - alpha)])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def WMA(self, N=10):
|
||||
|
|
@ -135,7 +137,7 @@ class Filter:
|
|||
w = np.arange(N, 0, -1)
|
||||
b = w / np.sum(w)
|
||||
a = np.array([1.0])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def MSMA(self, N=10):
|
||||
|
|
@ -149,7 +151,7 @@ class Filter:
|
|||
w[N] = 0.5
|
||||
b = w / N
|
||||
a = np.array([1.0])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def MLSQ(self, N=5):
|
||||
|
|
@ -170,11 +172,11 @@ class Filter:
|
|||
b = np.array([-11.0, 18.0, 88.0, 138.0, 168.0, 178.0, 168.0,
|
||||
138.0, 88.0, 18.0, -11.0]) / 980.0
|
||||
else:
|
||||
Y = self.X.copy()
|
||||
print("Warning: data returned unfiltered (wrong N)")
|
||||
self.idx = 0
|
||||
return Y
|
||||
return self.data
|
||||
a = np.array([1.0])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def ButterOrig(self, N=2, P=10):
|
||||
|
|
@ -196,10 +198,10 @@ class Filter:
|
|||
a = np.array([1.0, - (alpha + beta ** 2.0),
|
||||
(1.0 + alpha) * beta ** 2.0, - beta ** 4.0])
|
||||
else:
|
||||
Y = self.X.copy()
|
||||
print("Warning: data returned unfiltered (wrong N)")
|
||||
self.idx = 0
|
||||
return Y
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
return self.data
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def ButterMod(self, N=2, P=10):
|
||||
|
|
@ -220,10 +222,10 @@ class Filter:
|
|||
a = np.array([1.0, - (alpha + beta ** 2.0),
|
||||
(1.0 + alpha) * beta ** 2.0, - beta ** 4.0])
|
||||
else:
|
||||
Y = self.X.copy()
|
||||
print("Warning: data returned unfiltered (wrong N)")
|
||||
self.idx = 0
|
||||
return Y
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
return self.data
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def SuperSmooth(self, N=2, P=10):
|
||||
|
|
@ -246,10 +248,10 @@ class Filter:
|
|||
a = np.array([1.0, - (alpha + beta ** 2.0),
|
||||
(1.0 + alpha) * beta ** 2.0, - beta ** 4.0])
|
||||
else:
|
||||
Y = self.X.copy()
|
||||
print("Warning: data returned unfiltered (wrong N)")
|
||||
self.idx = 0
|
||||
return Y
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
return self.data
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def GaussLow(self, N=1, P=2):
|
||||
|
|
@ -259,16 +261,16 @@ class Filter:
|
|||
Must be P > 1. If not returns the unfiltered dataset.
|
||||
"""
|
||||
if (P < 2):
|
||||
Y = self.X.copy()
|
||||
print("Warning: data returned unfiltered (P < 2)")
|
||||
self.idx = 0
|
||||
return Y
|
||||
return self.data
|
||||
A = 2.0 ** (1.0 / N) - 1.0
|
||||
B = 4.0 * np.sin(np.pi / P) ** 2.0
|
||||
C = 2.0 * (np.cos(2.0 * np.pi / P) - 1.0)
|
||||
alpha = (-B + np.sqrt(B ** 2.0 - 4.0 * A * C)) / (2.0 * A)
|
||||
b = np.array([alpha])
|
||||
a = np.array([1.0, - (1.0 - alpha)])
|
||||
Y = self.X.copy()
|
||||
Y = self.data.copy()
|
||||
for i in range(N):
|
||||
Y, self.idx = filter_data(Y, b, a)
|
||||
return Y
|
||||
|
|
@ -280,16 +282,16 @@ class Filter:
|
|||
Must be P > 4. If not returns the unfiltered dataset.
|
||||
"""
|
||||
if (P < 5):
|
||||
Y = self.X.copy()
|
||||
print("Warning: data returned unfiltered (P < 5)")
|
||||
self.idx = 0
|
||||
return Y
|
||||
return self.data
|
||||
A = 2.0 ** (1.0 / N) * np.sin(np.pi / P) ** 2.0 - 1.0
|
||||
B = 2.0 * (2.0 ** (1.0 / N) - 1.0) * (np.cos(2.0 * np.pi / P) - 1.0)
|
||||
C = - B
|
||||
alpha = (-B - np.sqrt(B ** 2.0 - 4.0 * A * C)) / (2.0 * A)
|
||||
b = np.array([1.0 - alpha / 2.0, -(1.0 - alpha / 2.0)])
|
||||
a = np.array([1.0, - (1.0 - alpha)])
|
||||
Y = self.X - self.X[0, :]
|
||||
Y = self.data - self.data[0, :]
|
||||
for i in range(N):
|
||||
Y, self.idx = filter_data(Y, b, a)
|
||||
return Y
|
||||
|
|
@ -306,7 +308,7 @@ class Filter:
|
|||
alpha = 1.0 / gamma - np.sqrt(1.0 / gamma ** 2 - 1.0)
|
||||
b = np.array([(1.0 - alpha) / 2.0, 0.0, - (1.0 - alpha) / 2.0])
|
||||
a = np.array([1.0, - beta * (1.0 + alpha), alpha])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def BandStop(self, P=5, delta=0.3):
|
||||
|
|
@ -322,7 +324,7 @@ class Filter:
|
|||
b = np.array([(1.0 + alpha) / 2.0, - beta * (1.0 + alpha),
|
||||
(1.0 + alpha) / 2.0])
|
||||
a = np.array([1.0, -beta * (1.0 + alpha), alpha])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def ZEMA1(self, N=10, alpha=None, K=1.0, Vn=5):
|
||||
|
|
@ -337,7 +339,7 @@ class Filter:
|
|||
b[0] = alpha * (1.0 + K)
|
||||
b[Vn] = - alpha * K
|
||||
a = np.array([1.0, - (1.0 - alpha)])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def ZEMA2(self, N=10, alpha=None, K=1.0):
|
||||
|
|
@ -350,7 +352,7 @@ class Filter:
|
|||
alpha = 2.0 / (N + 1.0)
|
||||
b = np.array([alpha * (1.0 + K)])
|
||||
a = np.array([1.0, alpha * K - (1.0 - alpha)])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def InstTrend(self, N=10, alpha=None):
|
||||
|
|
@ -364,7 +366,7 @@ class Filter:
|
|||
b = np.array([alpha - alpha ** 2.0 / 4.0, alpha ** 2.0 / 2.0,
|
||||
- alpha + 3.0 * alpha ** 2.0 / 4.0])
|
||||
a = np.array([1.0, - 2.0 * (1.0 - alpha), (1.0 - alpha) ** 2.0])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def SincFunction(self, N=10, nel=10):
|
||||
|
|
@ -378,8 +380,8 @@ class Filter:
|
|||
k = np.arange(1, nel)
|
||||
b[1:] = np.sin(np.pi * k / N) / (np.pi * k)
|
||||
a = np.array([1.0])
|
||||
Y, self.idx = filter_data(self.X, b, a)
|
||||
return Y, b, a
|
||||
Y, self.idx = filter_data(self.data, b, a)
|
||||
return Y
|
||||
|
||||
def Decycler(self, P=10):
|
||||
"""
|
||||
|
|
@ -389,10 +391,10 @@ class Filter:
|
|||
Must be P > 4. If not returns the unfiltered dataset.
|
||||
"""
|
||||
if (P < 5):
|
||||
Y = self.X.copy()
|
||||
print("Warning: data returned unfiltered (P < 5)")
|
||||
self.idx = 0
|
||||
return Y
|
||||
Y = self.X - self.GaussHigh(N=1, P=P)
|
||||
return self.data
|
||||
Y = self.data - self.GaussHigh(N=1, P=P)
|
||||
return Y
|
||||
|
||||
def DecyclerOsc(self, P1=5, P2=10):
|
||||
|
|
@ -405,9 +407,111 @@ class Filter:
|
|||
"""
|
||||
P_low = np.amin([P1, P2])
|
||||
P_high = np.amax([P1, P2])
|
||||
if (P1 < 5):
|
||||
Y = self.X.copy()
|
||||
if (P_low < 5):
|
||||
print("Warning: data returned unfiltered (P_low < 5)")
|
||||
self.idx = 0
|
||||
return Y
|
||||
return self.data
|
||||
Y = self.GaussHigh(N=2, P=P_low) - self.GaussHigh(N=2, P=P_high)
|
||||
return Y
|
||||
|
||||
def ABG(self, alpha=0.0, beta=0.0, gamma=0.0, dt=1.0):
|
||||
"""
|
||||
alpha-beta-gamma
|
||||
|
||||
For numerical stability: 0 < alpha, beta < 1
|
||||
"""
|
||||
# If necessary change scalars to arrays
|
||||
if (np.ndim(alpha) == 0):
|
||||
alpha = np.ones(self.n_samples) * alpha
|
||||
if (np.ndim(beta) == 0):
|
||||
beta = np.ones(self.n_samples) * beta
|
||||
if (np.ndim(gamma) == 0):
|
||||
gamma = np.ones(self.n_samples) * gamma
|
||||
|
||||
# Initialize
|
||||
Y_corr = self.data.copy()
|
||||
Y_pred = self.data.copy()
|
||||
x0 = self.data[0,:]
|
||||
v0 = np.zeros(self.n_series)
|
||||
a0 = np.zeros(self.n_series)
|
||||
for i in range(1, self.n_samples):
|
||||
|
||||
# Predictor (predicts state in <i>)
|
||||
x_pred = x0 + dt * v0 + 0.5 * a0 * dt ** 2
|
||||
v_pred = v0 + dt * a0
|
||||
a_pred = a0
|
||||
Y_pred[i, :] = x_pred
|
||||
|
||||
# Residual (innovation)
|
||||
r = self.data[i, :] - x_pred
|
||||
|
||||
# Corrector (corrects state in <i>)
|
||||
x_corr = x_pred + alpha[i] * r
|
||||
v_corr = v_pred + (beta[i] / dt) * r
|
||||
a_corr = a_pred + (2.0 * gamma[i] / dt ** 2) *r
|
||||
|
||||
# Save value and prepare next iteration
|
||||
Y_corr[i, :] = x_corr
|
||||
x0 = x_corr
|
||||
v0 = v_corr
|
||||
a0 = a_corr
|
||||
|
||||
self.idx = 1
|
||||
|
||||
return Y_corr, Y_pred
|
||||
|
||||
|
||||
def Kalman(self, sigma_x, sigma_v, dt, abg_type="abg"):
|
||||
"""
|
||||
Steady-state Kalman filter (also limited to one-dimension)
|
||||
"""
|
||||
L = (sigma_x / sigma_v) * dt ** 2
|
||||
|
||||
# Alpha filter
|
||||
if (abg_type == 'a'):
|
||||
alpha = (-L ** 2 + np.sqrt(L ** 4 + 16.0 * L ** 2)) / 8.0
|
||||
beta = 0.0
|
||||
gamma = 0.0
|
||||
|
||||
# Alpha-Beta filter
|
||||
elif(abg_type == 'ab'):
|
||||
r = (4.0 + L - np.sqrt(8.0 * L + L ** 2)) / 4.0
|
||||
alpha = 1.0 - r ** 2
|
||||
beta = 2.0 * (2.0 - alpha) - 4.0 * np.sqrt(1.0 - alpha)
|
||||
gamma = 0.0
|
||||
|
||||
#Alpha-Beta-Gamma filter
|
||||
else:
|
||||
b = (L / 2.0) - 3.0
|
||||
c = (L / 2.0) + 3.0
|
||||
d = -1.0
|
||||
p = c - b **2 /3.0
|
||||
q = 2.0 * b **3 /27.0 - b * c /3.0 + d
|
||||
v = np.sqrt(q ** 2 + 4.0 * p ** 3 / 27.0)
|
||||
z = - (q + v / 2.0) ** (1.0 / 3.0)
|
||||
s = z - p / (3.0 * z) - b / 3.0
|
||||
alpha = 1.0 - s ** 2
|
||||
beta = 2.0 * (1 - s) ** 2
|
||||
gamma = beta ** 2 / (2.0 * alpha)
|
||||
|
||||
# Apply filter
|
||||
Y = self.abg(alpha=alpha, beta=beta, gamma=gamma, dt=dt)
|
||||
|
||||
return Y
|
||||
|
||||
"""
|
||||
correction = update = measurement
|
||||
prediction = motion
|
||||
|
||||
X (n_states, 1) State estimate
|
||||
P (n_states, n_states) Covariance estimate
|
||||
F (n_states, n_states) State transition model
|
||||
Z (n_obs, 1) Observations
|
||||
H (n_obs, n_states) Observation model
|
||||
R (n_obs, n_obs) Covariance of the observation noise
|
||||
S (n_obs, n_obs) Covariance of the observation residual
|
||||
K (n_states, n_obs) Optimal Kalman gain
|
||||
Q (n_states, n_states) Covariance of the process noise matrix
|
||||
Y (n_obs, 1) Observation residual (innovation)
|
||||
|
||||
"""
|
||||
|
|
@ -4,22 +4,22 @@ Filters for time series.
|
|||
Copyright (c) 2020 Gabriele Gilardi
|
||||
|
||||
ToDo:
|
||||
- use NaN/input values for points not filtered?
|
||||
- return idx?
|
||||
- util to test filter (impulse, utils)
|
||||
- warning in filter when wrong order? or save flag with true/false if computed
|
||||
- use self.a and self.b
|
||||
- remove a and b from plots
|
||||
- in comments write what filters do
|
||||
- is necessary to copy X for Y untouched?
|
||||
- decide default values in functions
|
||||
- check conditions on P and N
|
||||
- why lag plot gives errors
|
||||
- fix plotting function
|
||||
- example for alpha-beta-gamma using variable sigma as in financial time series
|
||||
(see Ehler)
|
||||
- example using noisy multi-sine-waves
|
||||
"""
|
||||
|
||||
import sys
|
||||
import numpy as np
|
||||
import filters as flt
|
||||
import utils as utl
|
||||
import matplotlib.pyplot as plt
|
||||
|
||||
# Read data to filter
|
||||
if len(sys.argv) != 2:
|
||||
|
|
@ -42,15 +42,19 @@ spx = flt.Filter(data)
|
|||
# aa = np.array([1.0, alpha - 1.0])
|
||||
|
||||
|
||||
res, bb, aa = spx.SincFunction(2, 50)
|
||||
print(bb)
|
||||
print(aa)
|
||||
utl.plot_frequency_response(bb, aa)
|
||||
utl.plot_lag_response(bb, aa)
|
||||
# res, bb, aa = spx.SincFunction(2, 50)
|
||||
# print(bb)
|
||||
# print(aa)
|
||||
# utl.plot_frequency_response(bb, aa)
|
||||
# utl.plot_lag_response(bb, aa)
|
||||
# sigma_x = 0.1
|
||||
# sigma_v = 0.1 * np.ones(n_samples)
|
||||
# res = spx.Kalman(sigma_x=sigma_x, sigma_v=sigma_v, dt=1.0, abg_type="abg")
|
||||
alpha = 0.5
|
||||
beta = 0.005
|
||||
gamma = 0.0
|
||||
Yc, Yp = spx.ABG(alpha=alpha, beta=beta, gamma=gamma, dt=1.0)
|
||||
signals = (spx.data[:, 0], Yc[:, 0], Yp[:, 0])
|
||||
utl.plot_signals(signals, 0, 50)
|
||||
|
||||
|
||||
# res = spx.DecyclerOsc(30, 60)
|
||||
# print(res[0:10, :])
|
||||
signals = (spx.X, res)
|
||||
print(spx.idx)
|
||||
utl.plot_signals(signals)
|
||||
# print(spx.X[0:20])
|
||||
|
|
@ -63,61 +63,20 @@ def scale_data(X, param=()):
|
|||
return Xs, param
|
||||
|
||||
|
||||
def calc_rmse(a, b):
|
||||
"""
|
||||
Calculates the root-mean-square-error of arrays <a> and <b>. If the arrays
|
||||
are multi-column, the RMSE is calculated as all the columns are one single
|
||||
vector.
|
||||
"""
|
||||
# Convert to (n, ) dimension
|
||||
a = a.flatten()
|
||||
b = b.flatten()
|
||||
|
||||
# Root-mean-square-error
|
||||
rmse = np.sqrt(((a - b) ** 2).sum() / len(a))
|
||||
|
||||
return rmse
|
||||
|
||||
|
||||
def calc_corr(a, b):
|
||||
"""
|
||||
Calculates the correlation between arrays <a> and <b>. If the arrays are
|
||||
multi-column, the correlation is calculated as all the columns are one
|
||||
single vector.
|
||||
"""
|
||||
# Convert to (n, ) dimension
|
||||
a = a.flatten()
|
||||
b = b.flatten()
|
||||
|
||||
# Correlation
|
||||
corr = np.corrcoef(a, b)[0, 1]
|
||||
|
||||
return corr
|
||||
|
||||
|
||||
def calc_accu(a, b):
|
||||
"""
|
||||
Calculates the accuracy (in %) between arrays <a> and <b>. The two arrays
|
||||
must be column/row vectors.
|
||||
"""
|
||||
# Convert to (n, ) dimension
|
||||
a = a.flatten()
|
||||
b = b.flatten()
|
||||
|
||||
# Correlation
|
||||
accu = 100.0 * (a == b).sum() / len(a)
|
||||
|
||||
return accu
|
||||
|
||||
|
||||
def plot_signals(signals):
|
||||
def plot_signals(signals, idx_start=0, idx_end=None):
|
||||
"""
|
||||
"""
|
||||
if (idx_end is None):
|
||||
idx_end = len(signals[0])
|
||||
t = np.arange(idx_start, idx_end)
|
||||
names = []
|
||||
count = 0
|
||||
for signal in signals:
|
||||
plt.plot(signal)
|
||||
|
||||
plt.plot(t, signal[idx_start:idx_end])
|
||||
names.append(str(count))
|
||||
count += 1
|
||||
plt.grid(b=True)
|
||||
plt.xlim(0, 100)
|
||||
plt.legend(names)
|
||||
plt.show()
|
||||
|
||||
|
||||
|
|
@ -138,8 +97,6 @@ def plot_frequency_response(b, a=1.0):
|
|||
# plt.ylim(-40.0, 0.0)
|
||||
plt.xlabel('$\omega$ [rad/sample]')
|
||||
plt.ylabel('$h$ [db]')
|
||||
plt.title('b = ' + np.array_str(np.around(b, decimals=2)) \
|
||||
+ ', a = ' + np.array_str(np.around(a, decimals=2)))
|
||||
plt.show()
|
||||
|
||||
|
||||
|
|
@ -157,128 +114,53 @@ def plot_lag_response(b, a=1.0):
|
|||
plt.xlim(np.amin(wf), np.amax(wf))
|
||||
plt.xlabel('$\omega$ [rad/sample]')
|
||||
plt.ylabel('$gd$ [samples]')
|
||||
plt.title('b = ' + np.array_str(np.around(b, decimals=2)) \
|
||||
+ ', a = ' + np.array_str(np.around(a, decimals=2)))
|
||||
plt.show()
|
||||
|
||||
|
||||
def synthetic_wave(per, amp=None, pha=None, tim=None):
|
||||
"""
|
||||
Generates a multi-sinewave.
|
||||
P = [ P1 P2 ... Pn ] Periods
|
||||
varargin = A, T, PH Amplitudes, time, phases
|
||||
A = [ A1 A2 ... An ] Amplitudes
|
||||
T = [ ts tf dt] Time info: from ts to tf in dt steps
|
||||
PH = [PH1 PH2 ... PHn] Phases (rad)
|
||||
|
||||
# function [y,t] = interpSignal(x,n1,n2,varargin)
|
||||
# if (nargin == 4)
|
||||
# dt = varargin{1};
|
||||
# else
|
||||
# dt = 20;
|
||||
# end
|
||||
# N = length(x);
|
||||
# Ts = (n2-n1)/(N-1);
|
||||
# n = n1:Ts:n2;
|
||||
# t = n1:Ts/dt:n2;
|
||||
# nt = length(t);
|
||||
# y = zeros(1,nt);
|
||||
# for i = 1:nt
|
||||
# a = x.*sinc( (t(i)-n)/Ts );
|
||||
# y(i) = sum(a);
|
||||
# end
|
||||
# end
|
||||
|
||||
|
||||
# function [f] = plotFn(type,func,varargin)
|
||||
# % 0 Plot real function
|
||||
# % 1 Plot real/imag values
|
||||
# % 2 Plot real/imag values in polar form
|
||||
# % 3 Plot magnitude/phase
|
||||
|
||||
# clf
|
||||
# tiny = 1e-7;
|
||||
|
||||
# % Check if func is a function or data
|
||||
# if ( isa(func,"function_handle") )
|
||||
# N = varargin{1};
|
||||
# n = (0:N-1)';
|
||||
# f = func(n);
|
||||
# else
|
||||
# N = length(func);
|
||||
# n = (0:N-1)';
|
||||
# f = func;
|
||||
# end
|
||||
|
||||
# % Clean data
|
||||
# xRe = real(f);
|
||||
# xIm = imag(f);
|
||||
# xRe( abs(xRe) < tiny ) = 0;
|
||||
# xIm( abs(xIm) < tiny ) = 0;
|
||||
|
||||
# switch (type)
|
||||
|
||||
# % Plot real function
|
||||
# case 0
|
||||
|
||||
# stem(n,xRe,"b","filled")
|
||||
# xlim([n(1) n(N)])
|
||||
# grid on
|
||||
# xlabel("n")
|
||||
# ylabel("f")
|
||||
# box on
|
||||
|
||||
# % Plot real/imag function
|
||||
# case 1
|
||||
# subplot(2,1,1)
|
||||
# stem(n,xRe,"b","filled")
|
||||
# xlim([n(1) n(N)])
|
||||
# grid on
|
||||
# xlabel("n")
|
||||
# ylabel("Re")
|
||||
# box on
|
||||
# subplot(2,1,2)
|
||||
# stem(n,xIm,"b","filled")
|
||||
# xlim([n(1) n(N)])
|
||||
# grid on
|
||||
# xlabel("n")
|
||||
# ylabel("Im")
|
||||
# box on
|
||||
|
||||
# % Plot real/imag function in polar form
|
||||
# case 2
|
||||
# scatter(xRe,xIm,"b","filled")
|
||||
# maxRe = max( abs(xRe) );
|
||||
# maxIm = max( abs(xIm) );
|
||||
# m = max(maxRe,maxIm);
|
||||
# dx = 2*m/50;
|
||||
# text(xRe+dx,xIm,num2str(n))
|
||||
# xlim( [-m +m ])
|
||||
# ylim( [-m +m ])
|
||||
# axis("square")
|
||||
# grid on
|
||||
# hold on
|
||||
# plot([-m 0; +m 0],[0 -m; 0 +m],"k")
|
||||
# hold off
|
||||
# xlabel("Real")
|
||||
# ylabel("Imag")
|
||||
# box on
|
||||
|
||||
# % Plot magnitude/phase
|
||||
# case 3
|
||||
# xMa = sqrt( xRe.^2 + xIm.^2 );
|
||||
# xAr = atan2(xIm,xRe);
|
||||
# subplot(2,1,1)
|
||||
# stem(n,xMa,"b","filled")
|
||||
# xlim([n(1) n(N)])
|
||||
# grid on
|
||||
# xlabel("n")
|
||||
# ylabel("Magnitude")
|
||||
# box on
|
||||
# subplot(2,1,2)
|
||||
# stem(n,xAr,"b","filled")
|
||||
# xlim([n(1) n(N)])
|
||||
# ylim([-pi pi])
|
||||
# grid on
|
||||
# xlabel("n")
|
||||
# ylabel("Phase [rad]")
|
||||
# box on
|
||||
Av = SUM(1 to n) of [ A*sin(2*pi*f*t + PH) ]
|
||||
Tv Time (ts to tf with step dt)
|
||||
|
||||
# end
|
||||
|
||||
# end
|
||||
Default amplitudes are ones
|
||||
Default time is from 0 to largest period (1000 steps)
|
||||
Default phases are zeros
|
||||
"""
|
||||
n_waves = len(per)
|
||||
per = np.asarray(per)
|
||||
|
||||
# Check for amplitudes, times, and phases
|
||||
if (amp is None):
|
||||
amp = np.ones(n_waves)
|
||||
else:
|
||||
amp = np.asarray(amp)
|
||||
if (tim is None):
|
||||
t_start = 0.0
|
||||
t_end = np.amax(per)
|
||||
n_steps = 500
|
||||
else:
|
||||
t_start = tim[0]
|
||||
t_end = tim[1]
|
||||
n_steps = int(tim[2])
|
||||
if (pha is None):
|
||||
pha = np.zeros(n_waves)
|
||||
else:
|
||||
pha = np.asarray(pha)
|
||||
|
||||
# Add all the waves
|
||||
t = np.linspace(t_start, t_end, num=n_steps)
|
||||
f = np.zeros(len(t))
|
||||
for i in range(n_waves):
|
||||
f = f + amp[i] * np.sin(2.0 * np.pi * t / per[i] + pha[i])
|
||||
|
||||
return t, f
|
||||
|
||||
|
||||
# function sP = SyntQT(P,type)
|
||||
|
|
@ -406,60 +288,5 @@ def plot_lag_response(b, a=1.0):
|
|||
# end % End of function
|
||||
|
||||
|
||||
# function [Tv,Av] = SyntWave(P,varargin)
|
||||
# % Function: generate a multi-sinewave
|
||||
# %
|
||||
# % Inputs
|
||||
# % ------
|
||||
# % P = [ P1 P2 ... Pn ] Periods
|
||||
# % varargin = A, T, PH Amplitudes, time, phases
|
||||
# %
|
||||
# % A = [ A1 A2 ... An ] Amplitudes
|
||||
# % T = [ ts tf dt] Time info: from ts to tf with step dt
|
||||
# % PH = [PH1 PH2 ... PHn] Phases (rad)
|
||||
# %
|
||||
# % Outputs
|
||||
# % -------
|
||||
# % Av = SUM(1 to n) of [ A*sin(2*pi*f*t + PH) ]
|
||||
# % Tv Time (ts to tf with step dt)
|
||||
# %
|
||||
# % Default amplitudes are ones
|
||||
# % Default time is from 0 to largest period (1000 steps)
|
||||
# % Default phases are zeros
|
||||
|
||||
# % Check for arguments
|
||||
# if (nargin == 1)
|
||||
# np = length(P);
|
||||
# A = ones(1,np);
|
||||
# T = [0 max(P) max(P)/1000];
|
||||
# PH = zeros(1,np);
|
||||
# elseif (nargin == 2)
|
||||
# np = length(P);
|
||||
# A = varargin{1};
|
||||
# T = [0 max(P) max(P)/1000];
|
||||
# PH = zeros(1,np);
|
||||
# elseif (nargin == 3)
|
||||
# np = length(P);
|
||||
# A = varargin{1};
|
||||
# T = varargin{2};
|
||||
# PH = zeros(1,np);
|
||||
# elseif (nargin == 4)
|
||||
# np = length(P);
|
||||
# A = varargin{1};
|
||||
# T = varargin{2};
|
||||
# PH = varargin{3};
|
||||
# else
|
||||
# fprintf(1,'\n');
|
||||
# error('Wrong number of arguments');
|
||||
# end
|
||||
|
||||
# % Add all waves
|
||||
# Tv = T(1):T(3):T(2);
|
||||
# Av = zeros(1,length(Tv));
|
||||
# for j = 1:np
|
||||
# Av = Av + A(j)*sin(2*pi*Tv/P(j)+PH(j));
|
||||
# end
|
||||
|
||||
# end % End of function
|
||||
|
||||
|
||||
|
|
|
|||
|
|
@ -2,9 +2,9 @@
|
|||
|
||||
## Reference
|
||||
|
||||
- [1] John F. Ehlers, "[Cycle Analytics for Traders: Advanced Technical Trading Concepts](http://www.mesasoftware.com/ehlers_books.htm)".
|
||||
- John F. Ehlers, "[Cycle Analytics for Traders: Advanced Technical Trading Concepts](http://www.mesasoftware.com/ehlers_books.htm)".
|
||||
|
||||
- [2] John F. Ehlers, "[Signal Analysis, Filters And Trading Strategies](http://www.mesasoftware.com/ehlers_technical_papers.htm)".
|
||||
- Wikipedia, "[Alpha beta filter](https://en.wikipedia.org/wiki/Alpha_beta_filter)".
|
||||
|
||||
## Characteristics
|
||||
|
||||
|
|
|
|||
Ładowanie…
Reference in New Issue