|
mi@0
|
1 #!/usr/bin/env python
|
|
mi@0
|
2 # encoding: utf-8
|
|
mi@0
|
3 """
|
|
mi@0
|
4 novelty.py
|
|
mi@0
|
5
|
|
mi@0
|
6 Created by mi tian on 2015-04-02.
|
|
mi@0
|
7 Copyright (c) 2015 __MyCompanyName__. All rights reserved.
|
|
mi@0
|
8 """
|
|
mi@0
|
9
|
|
mi@0
|
10 import sys, os
|
|
mi@0
|
11 import numpy as np
|
|
mi@0
|
12 from scipy.signal import correlate2d, convolve2d
|
|
mitian@1
|
13 import matplotlib.pyplot as plt
|
|
mitian@7
|
14 from utils.PeakPickerUtil import PeakPicker
|
|
mi@0
|
15
|
|
mitian@7
|
16
|
|
mitian@7
|
17 peak_picker = PeakPicker()
|
|
mitian@7
|
18 peak_picker.params.alpha = 9.0 # Alpha norm
|
|
mitian@7
|
19 peak_picker.params.delta = 0.0 # Adaptive thresholding delta
|
|
mitian@7
|
20 peak_picker.params.QuadThresh_a = (100 - 20.0) / 1000.0
|
|
mitian@7
|
21 peak_picker.params.QuadThresh_b = 0.0
|
|
mitian@7
|
22 peak_picker.params.QuadThresh_c = (100 - 20.0) / 1500.0
|
|
mitian@7
|
23 peak_picker.params.rawSensitivity = 20
|
|
mitian@7
|
24 peak_picker.params.aCoeffs = [1.0000, -0.5949, 0.2348]
|
|
mitian@7
|
25 peak_picker.params.bCoeffs = [0.1600, 0.3200, 0.1600]
|
|
mitian@7
|
26 peak_picker.params.preWin = 5
|
|
mitian@7
|
27 peak_picker.params.postWin = 5 + 1
|
|
mitian@7
|
28 peak_picker.params.LP_on = True
|
|
mitian@7
|
29 peak_picker.params.Medfilt_on = True
|
|
mitian@7
|
30 peak_picker.params.Polyfit_on = True
|
|
mitian@7
|
31 peak_picker.params.isMedianPositive = False
|
|
mitian@7
|
32
|
|
mi@0
|
33 def getNoveltyCurve(ssm, kernel_size, normalise=False):
|
|
mi@0
|
34 '''Return novelty score from ssm.'''
|
|
mi@0
|
35
|
|
mi@0
|
36 kernel_size = int(np.floor(kernel_size/2.0) + 1)
|
|
mi@0
|
37 stripe = getDiagonalSlice(ssm, kernel_size)
|
|
mi@0
|
38 kernel = gaussian_kernel(kernel_size)
|
|
mi@0
|
39 xc = convolve2d(stripe,kernel,mode='same')
|
|
mi@0
|
40 xc[abs(xc)>1e+10]=0.00001
|
|
mi@0
|
41
|
|
mi@0
|
42 novelty = xc[int(np.floor(xc.shape[0]/2.0)),:]
|
|
mi@0
|
43 novelty = [0.0 if (np.isnan(x) or np.isinf(x) or x > 1e+100) else x for x in novelty]
|
|
mi@0
|
44
|
|
mi@0
|
45 if normalise:
|
|
mi@0
|
46 novelty = (novelty - np.min(novelty)) / (np.max(novelty) - np.min(novelty))
|
|
mi@0
|
47 return novelty
|
|
mi@0
|
48
|
|
mitian@8
|
49 def getNoveltyDiff(novelty, N=1, relative=False):
|
|
mitian@8
|
50 '''Return the second order differece in the novelty curve.'''
|
|
mitian@8
|
51
|
|
mitian@8
|
52 diff = np.zeros_like(novelty)
|
|
mitian@8
|
53 diff[:-N] = np.diff(novelty, n=N)
|
|
mitian@8
|
54
|
|
mitian@8
|
55 if relative:
|
|
mitian@8
|
56 diff /= novelty
|
|
mitian@8
|
57
|
|
mitian@8
|
58 return diff
|
|
mitian@8
|
59
|
|
mi@0
|
60 def getDiagonalSlice(ssm, width):
|
|
mi@0
|
61 ''' Return a diagonal stripe of the ssm given its width, with 45 degrees rotation.
|
|
mi@0
|
62 Note: requres 45 degrees rotated kernel also.'''
|
|
mi@0
|
63 w = int(np.floor(width/2.0))
|
|
mi@0
|
64 length = len(np.diagonal(ssm))
|
|
mi@0
|
65 stripe = np.zeros((2*w+1,length))
|
|
mi@0
|
66 # print 'diagonal', length, w, stripe.shape
|
|
mi@0
|
67 for i in xrange(-w, w+1) :
|
|
mi@0
|
68 stripe[w+i,:] = np.hstack(( np.zeros(int(np.floor(abs(i)/2.0))), np.diagonal(ssm,i), np.zeros(int(np.ceil(abs(i)/2.0))) ))
|
|
mi@0
|
69 return stripe
|
|
mi@0
|
70
|
|
mi@0
|
71 def gaussian_kernel(size):
|
|
mi@0
|
72 '''Create a gaussian tapered 45 degrees rotated checkerboard kernel.
|
|
mi@0
|
73 TODO: Unit testing: Should produce this with kernel size 3:
|
|
mi@0
|
74 0.1353 -0.3679 0.1353
|
|
mi@0
|
75 0.3679 1.0000 0.3679
|
|
mi@0
|
76 0.1353 -0.3679 0.1353
|
|
mi@0
|
77 '''
|
|
mi@0
|
78 n = float(np.ceil(size / 2.0))
|
|
mi@0
|
79 kernel = np.zeros((size,size))
|
|
mi@0
|
80 for i in xrange(1,size+1) :
|
|
mi@0
|
81 for j in xrange(1,size+1) :
|
|
mi@0
|
82 gauss = np.exp( -4.0 * (np.square( (i-n)/n ) + np.square( (j-n)/n )) )
|
|
mi@0
|
83 # gauss = 1
|
|
mi@0
|
84 if np.logical_xor( j - n > np.floor((i-n) / 2.0), j - n > np.floor((n-i) / 2.0) ) :
|
|
mi@0
|
85 kernel[i-1,j-1] = -gauss
|
|
mi@0
|
86 else:
|
|
mi@0
|
87 kernel[i-1,j-1] = gauss
|
|
mi@0
|
88
|
|
mi@0
|
89 return kernel
|
|
mi@0
|
90
|
|
mitian@17
|
91 def process(ssm, peak_picker=peak_picker, kernel_size=48, normalise=False, plot=False):
|
|
mi@0
|
92 '''Detect segment boundaries in the ssm.'''
|
|
mitian@7
|
93 # peak_picker for the 1st round boudary detection
|
|
mitian@7
|
94
|
|
mitian@7
|
95
|
|
mi@0
|
96 novelty = getNoveltyCurve(ssm, kernel_size, normalise=False)
|
|
mi@0
|
97 smoothed_novelty, novelty_peaks = peak_picker.process(novelty)
|
|
mi@0
|
98
|
|
mitian@1
|
99 if plot:
|
|
mitian@1
|
100 plot_detection(smoothed_novelty, novelty_peaks)
|
|
mitian@1
|
101 return novelty, smoothed_novelty, novelty_peaks
|
|
mitian@1
|
102
|
|
mitian@1
|
103 def plot_detection(smoothed_novelty, novelty_peaks):
|
|
mitian@17
|
104 pass
|
