Mercurial > hg > camir-aes2014

comparison toolboxes/MIRtoolbox1.3.2/MIRToolboxDemos/demo4segmentation.m @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:e9a9cd732c1e
1 function demo4segmentation
2 % To get familiar with some approaches of segmentation of audio files
3 % using MIRtoolbox.
4
5 % 1. Load an audio file (for instance, guitar.wav).
6 a = miraudio('guitar');
7
8 % 2. We will perform the segmentation strategy as proposed in (Foote &
9 % Cooper, 2003). First, decompose the file into successive frames of 50 ms
10 % without overlap.
11 help mirframe
12 fr = mirframe(a,0.05,1)
13
14 % 3. Compute the spectrum representation (FFT) of the frames.
15 sp = mirspectrum(fr)
16 clear fr
17 % (Remove from the memory any data that will not be used any more.)
18
19 % 4. Compute the similarity matrix that shows the similarity between the
20 % spectrum of different frames.
21 help mirsimatrix
22 sm = mirsimatrix(sp)
23 clear sp
24 % Look at the structures shown in the matrix and find the relation with the
25 % structure heard when listening to the extract.
26
27 % 5. Estimate the novelty score related to the similarity matrix. It
28 % consists in a convolution of the diagonal of the matrix with a
29 % checker-board Gaussian kernel. Use the novelty function for that purpose.
30 help mirnovelty
31 nv = mirnovelty(sm)
32
33 % 6. Detect the peaks in the novelty score.
34 help mirpeaks
35 p1 = mirpeaks(nv)
36
37 % You can change the threshold value of the peak picker function in order to
38 % get better results.
39 p2 = mirpeaks(nv,'Contrast',0.01)
40
41 clear nv
42
43 % 7. Segment the original audio file using the peaks as position for
44 % segmentation.
45 help mirsegment
46 s1 = mirsegment(a,p1)
47 clear p1
48
49 % 8. Listen to the results.
50 mirplay(s1)
51
52 %s2 = mirsegment(a,p2)
53 %clear p2
54 %mirplay(s2)
55
56 % 9. Compute the similarity matrix of this obtained segmentation, in order
57 % to view the relationships between the different segments and their
58 % possible clustering into higher-level groups.
59 mirsimatrix(s1,'Similarity')
60 clear s1
61 %mirsimatrix(s2)
62 %clear s2
63
64 display('Strike any key to continue...');
65 pause
66 close all
67
68 % 10. Change the size of the kernel used in the novelty function, in order
69 % to obtain segmentations of different levels of detail, from detailed
70 % analysis of the local texture, to very simple segmentation of the whole
71 % piece.
72 n100 = mirnovelty(sm,'KernelSize',100)
73 n50 = mirnovelty(sm,'KernelSize',50)
74 n10 = mirnovelty(sm,'KernelSize',10)
75 clear sm
76 % As you can see, the smaller the gaussian kernel is, the more peaks can be
77 % found in the novelty score. Indeed, if the kernel is small, the cumulative
78 % multiplication of its elements with the superposed elements in the
79 % similarity matrix may vary more easily, throughout the progressive
80 % sliding of the kernel along the diagonal of the similarity matrix, and
81 % local change of texture may be more easily detected. On the contrary,
82 % when the kernel is large, only large-scale change of texture are
83 % detected.
84
85 display('Strike any key to continue...');
86 pause
87 close all
88
89 p100 = mirpeaks(n100,'NoBegin','NoEnd')
90 clear n100
91 p50 = mirpeaks(n50,'NoBegin','NoEnd')
92 clear n50
93 p10 = mirpeaks(n10,'NoBegin','NoEnd')
94 clear n10
95 s100 = mirsegment(a,p100)
96 clear p100
97 mirplay(s100)
98 clear s100
99 s50 = mirsegment(a,p50)
100 clear p50
101 mirplay(s50)
102 clear s50
103 s10 = mirsegment(a,p10)
104 clear p10
105 mirplay(s10)
106 clear s10
107
108 display('Strike any key to continue...');
109 pause
110 close all
111
112 % One more compact way of writing these commands is as follows:
113 mirsegment(a,'Novelty')
114 mirsegment(a,'Novelty','Contrast',0.01)
115 mirsegment(a,'Novelty','KernelSize',100)
116
117 display('Strike any key to continue...');
118 pause
119 close all
120
121 % Besides, if you want to see the novelty curve with the peaks, just add a
122 % second output:
123 [s50 p50] = mirsegment(a,'Novelty','KernelSize',50)
124 clear s50 p50
125 [s10 p10] = mirsegment(a,'Novelty','KernelSize',10)
126 clear a s10 p10
127
128 display('Strike any key to continue...');
129 pause
130 close all
131
132
133 % 11. Try the whole process with MFCC instead of spectrum analysis. Take the
134 % first ten MFCC for instance.
135 help mirsegment
136 % The segment function can simply be called as follows:
137 sc = mirsegment('czardas','Novelty','MFCC','Rank',1:10)
138 clear sc
139
140 % Here are some other examples of use:
141 [ssp p m b] = mirsegment('valse_triste_happy','Spectrum',...
142 'KernelSize',150,'Contrast',.1)
143 clear p m b
144 mirplay(ssp)
145 clear ssp
146
147 display('Strike any key to continue...');
148 pause
149 close all
150
151 [smfcc2 p m a] = mirsegment('valse_triste_happy','MFCC',2:10,...
152 'KernelSize',150,'Contrast',.1)
153 clear p m a
154 mirplay(smfcc2)