-
Notifications
You must be signed in to change notification settings - Fork 7
/
Copy pathsynthetic_data.py
210 lines (204 loc) · 9.21 KB
/
synthetic_data.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
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
import numpy as np
import matplotlib.pyplot as plt
nsamples = 64
train_cnt = 3000
maxbw=30
minbw=7
fcnt=6
scnt = nsamples * fcnt
predict=False
#ltraffic=["single_cont", "mult_cont", "single_rshort", "mult_rshort", "det_hop"]
ltraffic=["single_cont", "single_rshort", "mult_cont", "det_hop"]
#ltraffic=["single_cont", "single_rshort", "det_hop"]
dl=[]
for el in ltraffic:
dl += [el]+["dummy"]*2
#ltraffic=dl
noiseval = 0.01
mindb=5
maxdb=20
plotenable=False
datatype = "float32"
train_data = np.zeros((1,fcnt,nsamples),dtype=datatype)
train_labels = np.zeros((1,len(ltraffic)),dtype=datatype)
bw_labels = []
pos_labels = []
nsig_labels = []
def gaussian(cnt):
indata = noiseval/np.sqrt(2)* (np.random.normal(size=(cnt,fcnt,nsamples)) + 1j* np.random.normal(size=(cnt,fcnt,nsamples)))
outdata = 20*np.log10(np.abs(np.fft.fft(indata)/float(nsamples)))
outdata = np.fft.fftshift(outdata,axes=(2,))
#outdata = -np.min(outdata)+outdata
#outdata = (outdata- np.min(outdata))/(1 - np.min(outdata))
return outdata
if predict:
pred_data = np.zeros((1,pcnt,nsamples),dtype=datatype)
def gendata(noise=True, normalize=True):
global train_data, train_labels, bw_labels, pos_labels, nsig_labels
idx=0
for traffic in ltraffic:
if traffic == "single_cont":
for i in range(train_cnt):
sigbw = np.random.randint(minbw,maxbw)
sigpos = np.random.randint(0, nsamples-sigbw)
bw_labels.append((sigbw/float(nsamples)))
pos_labels.append((sigpos+(sigbw/2.0))/float(nsamples))
nsig_labels.append(1)
sigon1=np.concatenate((np.ones((fcnt,sigbw)),np.zeros((fcnt,nsamples-sigbw))),axis=1)
sigon1 = np.roll(sigon1,sigpos,axis=1).reshape(1,fcnt,nsamples)
snr = np.random.uniform(mindb,maxdb)
rintens = 10**(snr/20.0) * noiseval * nsamples/sigbw
sigon1 = sigon1 * rintens #+ gaussian()
train_data = np.vstack((train_data,sigon1))
dlabels=np.zeros((train_cnt,len(ltraffic)))
dlabels[:,idx]=1
train_labels = np.vstack((train_labels,dlabels))
if plotenable:
plt.figure(1)
print pos_labels[0]*nsamples
print bw_labels[0]*nsamples
plt.imshow(sigon1[0], interpolation='none')
plt.colorbar(orientation='vertical')
plt.show()
elif traffic == "mult_cont":
for i in range(train_cnt):
nsig = np.random.randint(2,5)
actsig=np.zeros((fcnt,nsamples))
posum=0
for i in range(nsig):
sigbw = np.random.randint(minbw,maxbw/2)
sigpos = np.random.randint(0, nsamples-sigbw)
posum += sigpos+(sigbw/2.0)
sigon1=np.concatenate((np.ones((fcnt,sigbw)),np.zeros((fcnt,nsamples-sigbw))),axis=1)
sigon1 = np.roll(sigon1,sigpos,axis=1).reshape(1,fcnt,nsamples)
snr = np.random.uniform(mindb,maxdb)
rintens = 10**(snr/20.0) * noiseval * nsamples/sigbw
sigon1 = sigon1 * rintens
actsig = actsig + sigon1
orgbw = len(np.where(actsig[0]>0)[0])
bw_labels.append((orgbw/float(nsamples)))
pos_labels.append(posum/float(nsamples)/nsig)
nsig_labels.append(nsig)
actsig = actsig #+ gaussian()
train_data = np.vstack((train_data,actsig))
dlabels=np.zeros((train_cnt,len(ltraffic)))
dlabels[:,idx]=1
train_labels = np.vstack((train_labels,dlabels))
if plotenable:
plt.figure(1)
plt.imshow(actsig[0], interpolation='none')
plt.colorbar(orientation='vertical')
plt.show()
elif traffic == "single_rshort":
for i in range(train_cnt):
sigbw = np.random.randint(minbw,maxbw)
sigpos = np.random.randint(0, nsamples-sigbw)
bw_labels.append((sigbw/float(nsamples)))
pos_labels.append((sigpos+(sigbw/2.0))/float(nsamples))
nsig_labels.append(1)
ipres = np.random.randint(2,size=fcnt).reshape(fcnt,1)
sig = np.repeat(ipres,sigbw,axis=1)
sigon1=np.zeros((fcnt,nsamples))
sigon1[:,sigpos:sigpos+sigbw] = sig
snr = np.random.uniform(mindb,maxdb)
rintens = 10**(snr/20.0) * noiseval * nsamples/sigbw
sigon1 = sigon1 * rintens #+ gaussian()
sigon1= sigon1.reshape(1,fcnt,nsamples)
train_data = np.vstack((train_data,sigon1))
dlabels=np.zeros((train_cnt,len(ltraffic)))
dlabels[:,idx]=1
train_labels = np.vstack((train_labels,dlabels))
if plotenable:
plt.figure(1)
plt.imshow(sigon1[0], interpolation='none')
plt.colorbar(orientation='vertical')
plt.show()
elif traffic == "mult_rshort":
for i in range(train_cnt):
nsig = np.random.randint(2,5)
actsig=np.zeros((fcnt,nsamples))
posum=0
for i in range(nsig):
sigbw = np.random.randint(minbw,maxbw/2)
sigpos = np.random.randint(0, nsamples-sigbw)
posum += sigpos+(sigbw/2.0)
ipres = np.random.randint(2,size=fcnt).reshape(fcnt,1)
sig = np.repeat(ipres,sigbw,axis=1)
sigon1=np.zeros((fcnt,nsamples))
sigon1[:,sigpos:sigpos+sigbw] = sig
snr = np.random.uniform(mindb,maxdb)
rintens = 10**(snr/20.0) * noiseval * nsamples/sigbw
sigon1 = sigon1 * rintens
actsig = actsig + sigon1
actsig = actsig.reshape(1,fcnt,nsamples)
#maxpos = np.argmax(np.sum(actsig>0,axis=0))
#orgbw = len(np.where(actsig[0][maxpos]>0)[0])
orgbw = np.max(np.sum(actsig>0,axis=0))
bw_labels.append((orgbw/float(nsamples)))
pos_labels.append(posum/float(nsamples)/nsig)
nsig_labels.append(nsig)
actsig = actsig #+ gaussian()
train_data = np.vstack((train_data,actsig))
dlabels=np.zeros((train_cnt,len(ltraffic)))
dlabels[:,idx]=1
train_labels = np.vstack((train_labels,dlabels))
if plotenable:
plt.figure(1)
plt.imshow(actsig[0], interpolation='none')
plt.colorbar(orientation='vertical')
plt.show()
elif traffic == "det_hop":
for i in range(train_cnt):
sigbw = np.random.randint(minbw,maxbw/2)
sigpos = np.random.randint(0, nsamples-sigbw)
detshift = np.random.randint(1,maxbw/4)
initarr=np.zeros(nsamples)
initarr[sigpos:sigpos+sigbw] = 1
res=[]
for j in range(fcnt):
res.append(initarr)
initarr= np.roll(initarr,detshift)
res = np.reshape(res,(1,fcnt,nsamples))
#orgbw = sigbw+detshift*fcnt
orgbw = sigbw
bw_labels.append((orgbw/float(nsamples)))
pos_labels.append((sigpos+(sigbw/2.0)+(fcnt*detshift/2))%nsamples/float(nsamples))
snr = np.random.uniform(mindb,maxdb)
rintens = 10**(snr/20.0) * noiseval * nsamples/sigbw
res = res * rintens #+ gaussian()
train_data = np.vstack((train_data,res))
nsig_labels.append(1)
dlabels=np.zeros((train_cnt,len(ltraffic)))
dlabels[:,idx]=1
train_labels = np.vstack((train_labels,dlabels))
if plotenable:
plt.figure(1)
plt.imshow(res[0], interpolation='none')
plt.colorbar(orientation='vertical')
plt.show()
else:
print("Traffic not defined")
idx+=1
train_data = np.delete(train_data,0,0)
train_labels = np.delete(train_labels,0,0)
bw_labels = np.reshape(bw_labels,(-1,1))
pos_labels = np.reshape(pos_labels,(-1,1))
nsig_labels = np.reshape(nsig_labels,(-1,1))
if predict:
pred_data = np.delete(pred_data,0,0)
print("Predict data:",pred_data.shape)
print("--"*50)
print("Training data:",train_data.shape)
print("--"*50)
#print len(train_data), len(train_labels), len(bw_labels), len(pos_labels)
if noise:
#train_data_nn = train_data
#train_data = train_data+gaussian(train_data.shape[0])
train_data = train_data+ noiseval
train_data_org = np.copy(train_data)
train_data = 20 * np.log10(train_data)
nmin = np.min(train_data)
train_data = -np.min(train_data)+train_data
nmax = np.max(train_data)
train_data = (train_data- np.min(train_data))/(np.max(train_data) - np.min(train_data))
return train_data, train_labels, bw_labels, pos_labels, nmin, nmax, train_data_org