-
Notifications
You must be signed in to change notification settings - Fork 0
/
signal_provider.py
189 lines (151 loc) · 6.21 KB
/
signal_provider.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
# -*- coding: utf-8 -*-
from __future__ import division, print_function, absolute_import
import numpy as np
import matplotlib.pyplot as plt
import random
import os
_signal_type = 0
_data_dir = 'signal-rec'
_custom_data_fn = os.path.join(_data_dir, 'tracking_data.npz')
_sample_rate = 100
_training_times = 50
_testing_times = 10
def sweep(sp):
k = 0.0187 * (np.exp(4 * sp / (_training_times + _testing_times)) - 1)
w = 0.1 + k * (2 - 0.1)
return np.sin(w * sp)
def signal_fun(sp):
switcher = {
0: np.sin,
1: sweep,
}
f = switcher.get(_signal_type)
return f(sp)
def generate_signal():
ls = np.linspace(0, _training_times, _training_times * _sample_rate, dtype=np.float32)
f1 = signal_fun(ls)
ls = np.linspace(_training_times, _training_times + _testing_times, _testing_times * _sample_rate,
dtype=np.float32)
f2 = signal_fun(ls)
f1 = f1 + wgn(f1, 50)
return f1, f2
def wgn(x, snr):
snr = 10 ** (snr / 10.0)
x_power = np.sum(x ** 2) / len(x)
n_power = x_power / snr
return np.random.randn(len(x)) * np.sqrt(n_power)
def generate_data(seq, in_steps, out_steps):
xs = []
ys = []
if seq.ndim == 3:
for d in range(seq.shape[0]):
seq1 = seq[d]
for i in range(len(seq1) - in_steps - out_steps + 1):
xs.append([seq1[i: i+in_steps]])
ys.append([seq1[i+in_steps: i+in_steps+out_steps]])
else:
for i in range(len(seq) - in_steps - out_steps + 1):
xs.append([seq[i: i+in_steps]])
ys.append([seq[i+in_steps: i+in_steps+out_steps]])
return np.array(xs, dtype=np.float32), np.array(ys, dtype=np.float32)
class SignalProvider(object):
def __init__(self, batch_size=32, custom_data=True, sample_gap=1, shuffle=True, refresh_data=False,
input_steps=20, predict_steps=1, predict_depth=1):
if not os.path.exists(_data_dir):
os.mkdir(_data_dir)
if custom_data and _signal_type != 9:
custom_data = False
print('[WARNING]: signal_type != 9, USE SELF GENERATE DATA!')
self._signal_fn = os.path.join(_data_dir, 'signal-t{}.npz'.format(_signal_type))
self._predict_depth = predict_depth
if os.path.isfile(self._signal_fn):
data = np.load(self._signal_fn)
self._x = data['x']
self._y = data['y']
if self._x.shape[1] != input_steps or self._y.shape[1] != predict_steps:
refresh_data = True
self._evaluate_dat = data['eval']
self.norm_mean = data['norm_mean']
self.norm_std = data['norm_std']
else:
refresh_data = True
# re-sampling the data if needed
if refresh_data:
if custom_data:
data = np.load(_custom_data_fn)
f1 = data['x']
f2 = data['y']
else:
f1, f2 = generate_signal()
f1 = np.expand_dims(np.expand_dims(f1, axis=0), axis=-1)
f2 = np.expand_dims(np.expand_dims(f2, axis=0), axis=-1)
print('[DATA-INFO]: Train data shape: {} (nSeqs, nSteps, nFeats)'.format(f1.shape))
print('[DATA-INFO]: Test data shape: {} (nSeqs, nSteps, nFeats)'.format(f2.shape))
depth = f1.shape[2]
self.norm_mean = np.mean(np.reshape(f1, [-1, depth]), axis=0)
self.norm_std = np.std(np.reshape(f1, [-1, depth]), axis=0)
f1 = (f1 - self.norm_mean) / self.norm_std
print('[DATA-INFO]: Train data mean: {}, std: {}'.format(self.norm_mean, self.norm_std))
self._x, self._y = generate_data(f1, input_steps, predict_steps)
self._x = np.squeeze(self._x, axis=1)
self._y = np.squeeze(self._y, axis=1)
self._evaluate_dat = f2
np.savez(self._signal_fn, x=self._x, y=self._y, eval=self._evaluate_dat,
norm_mean=self.norm_mean, norm_std=self.norm_std)
self._batch_size = batch_size
self._sample_gap = sample_gap
self._shuffle = shuffle
self._fetch_id = 0
self._sample_len = self._x.shape[0]
self._index_arr = np.arange(self._sample_len)
self._prep_steps = 200
if self._shuffle:
np.random.shuffle(self._index_arr)
def get_next_batch(self):
x_ids = self._index_arr[self._fetch_id:self._fetch_id + self._batch_size]
x_ = self._x[x_ids, :]
y_ = self._y[x_ids, :]
self._fetch_id += self._sample_gap
if self._fetch_id + self._batch_size > self._sample_len:
if self._shuffle:
np.random.shuffle(self._index_arr)
self._fetch_id = 0
return x_, y_, self._fetch_id
def evaluate_length(self):
return len(self._evaluate_dat[0]) - self._prep_steps
def evaluate_dat(self):
return self._evaluate_dat[:, 0:self._prep_steps]
def evaluate(self, predicted):
''''''
e, = plt.plot(predicted[0, :], label='signal_e')
g, = plt.plot(self._evaluate_dat[0, self._prep_steps:], label='signal_gt')
plt.legend([e, g], ['predicted', 'gt'])
plt.show()
''''''
# calculate root mean square error
return np.sqrt(((predicted - self._evaluate_dat[:, self._prep_steps:, :self._predict_depth]) ** 2).mean())
def evaluate_dat_v2(self):
l = self.evaluate_length()
tx = []
for d in range(self._evaluate_dat.shape[0]):
tx1 = []
for i in range(l):
d1 = self._evaluate_dat[d, i:i + self._prep_steps]
tx1.append(d1)
tx.append(np.array(tx1, dtype=np.float32))
return np.array(tx, dtype=np.float32)
if __name__ == '__main__':
sp = SignalProvider()
dat = sp.evaluate_dat_v2()
dat = dat[:,:,-1]
pv = dat[:,1:,:1]-dat[:,:-1,:1]
pv = np.concatenate((pv[:,:1,:], pv), axis=1)
predicted = dat[:,:,:1] + pv # +dat[:,:,1:2]+dat[:,:,2:3]
# pa = (pa[:,:-1]*0.1+pa[:,1:]*0.9)
# pa = np.concatenate((p0, pa), axis=1)
# plt.plot(dat[0,:,2], label='signal_gt')
# plt.show()
print(sp.evaluate(predicted))
# f1, f2 = generate_signal()
# plt.plot(f1, label='signal_gt')
# plt.show()