-
Notifications
You must be signed in to change notification settings - Fork 0
/
functions_denoise.py
151 lines (114 loc) · 4.68 KB
/
functions_denoise.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
import numpy as np
from reading_dataset import *
from ecg2vcg import leads2vcg, limb2augmented
import statistics
import math
import scipy
from scipy.signal import butter, filtfilt
from denoise_wavelet import denoising_data
from denoise_emd import denoising_data_emd
import wfdb
def lowpass (signal):
fs=500
cutoff=10
nyq=0.15*fs
order=5
normal_cutoff = cutoff / nyq
b, a = butter(order, normal_cutoff, btype='low', analog=False)
y = filtfilt(b, a, signal)
return y
def lowpass_forall (mat):
a=np.zeros(mat.shape)
for i in range (len (mat)):
denoised=lowpass(mat[i])
a[i,:]=denoised
return a
def clean_the_ecg (mat,method='lowpass'):
if method=='lowpass':
clean_mat=lowpass_forall(mat)
return (clean_mat)
def calculate_snr_percent(signal, noise):
signal_power = np.sum(signal ** 2) / len(signal)
noise_power = np.sum(noise ** 2) / len(noise)
snr = (signal_power / noise_power)
return snr
def calculate_snr(signal, noise):
signal_power = np.sum(signal ** 2) / len(signal)
noise_power = np.sum(noise ** 2) / len(noise)
snr = 10 * np.log10(signal_power / noise_power)
return snr
def calculate_noise_signal(clean_signal, noise_signal, snr_db):
# Calculate the power of the clean signal
clean_power = np.sum(clean_signal ** 2) / len(clean_signal)
# Calculate the power of the noise signal
noise_power = np.sum(noise_signal ** 2) / len(noise_signal)
# Calculate the power of the desired noise signal
desired_noise_power = clean_power / (10 ** (snr_db / 10))
# Calculate the coefficient
coefficient = np.sqrt(desired_noise_power / noise_power)
final_noise=coefficient*(noise_signal)
return final_noise
def create_noise (mat , noise_method,snr_db=6 ,percent_of_file=1 ):
if snr_db==None:
return np.zeros(np.shape(mat))
else:
mat=mat/1000
length=len(mat[0])
# noisy_mat=np.zeros((np.shape(mat)[0],np.shape(mat)[1]))
a=np.shape(mat)
noise_mat=np.zeros(a)
thetas=[0,-np.pi/3,-np.pi*2/3 ,5*np.pi/6,np.pi/6,-np.pi/2]
x=np.arange(0,length)
if noise_method =='Artificial_bw_noise':
a1=np.random.uniform(low=0.1,high=0.5)
a2=np.random.uniform(low=0.1,high=0.3)
b1=np.random.uniform(low=0.1,high=0.8)/1000
b2=np.random.uniform(low=0.1,high=0.8)/1000
a1=0.33
a2=0.22
# y= (a1*np.sin(b1*2*np.pi*x)+ a2*np.cos(b2*2*np.pi*x))*1000
final_noise=(0.33*np.sin(0.0004*2*np.pi*x)+0.22*np.cos(0.0009*2*np.pi*x))
elif noise_method=='wn' :
final_noise=0.05*np.random.randn(mat[0].size)
final_noise=calculate_noise_signal(mat[0],final_noise,snr_db=snr_db)
else:
# rand=np.random.uniform(0,percent_of_file)
# noisy_amount = 10000*percent_of_file
# rand = (10000-noisy_amount)
loc='dataset\\mit-bih-noise-stress-test-database-1.0.0\\'
noise_file = loc +noise_method
noises, fields = wfdb.rdsamp(noise_file, sampfrom=0 , sampto=5000)
noise=noises[:,0]
if snr_db!='whole':
final_noise=calculate_noise_signal(mat[0],noise,snr_db=snr_db)
else:
final_noise=noise
for i in range(len(thetas)) :
noise_mat[i]=final_noise*np.cos(thetas[i])
# plt.figure()
# plt.plot(mat[i])
# plt.plot(noisy_mat[i])
return noise_mat *1000
def denoise_wavelet (noisy_mat ,wavelet_type,wavelet_denoise_treshold=0.13):
if wavelet_type=='just_baseline':
denoised_mat = denoising_data(noisy_mat,wavelet_type=wavelet_type,wavelet_denoise_treshold=wavelet_denoise_treshold,
just_baseline=True)
else :
denoised_mat = denoising_data(noisy_mat,wavelet_type=wavelet_type,wavelet_denoise_treshold=wavelet_denoise_treshold,
just_baseline=False)
return denoised_mat
# from skimage import restoration
# def denoise_wavelet2(mat , method ='BayesShrink', mode='soft' , wavelet_type='sym8'):
# denoised_list=[]
# for i in range (6) :
# noisy_signal=mat[i]
# denoised_signal_bayes=restoration.denoise_wavelet(noisy_signal, method =method, mode=mode , wavelet=wavelet_type,rescale_sigma='True')
# denoised_list.append(denoised_signal_bayes)
# denoised_mat= np.array(denoised_list)
# return denoised_mat
def denoise_emd_baseline(noisy_mat, cemd=False ,sd_tresh=0.05):
denoisedmat=denoising_data_emd (noisy_mat,sd_tresh=sd_tresh )
if cemd== False :
return denoisedmat[0]
else:
return denoisedmat[1]