cannam@24
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
cannam@24
|
2
|
cannam@18
|
3 /*
|
cannam@24
|
4 * ClusterMeltSegmenter.h
|
cannam@18
|
5 *
|
cannam@24
|
6 * Created by Mark Levy on 23/03/2006.
|
cannam@24
|
7 * Copyright 2006 Centre for Digital Music, Queen Mary, University of London.
|
cannam@24
|
8 * All rights reserved.
|
cannam@18
|
9 */
|
cannam@18
|
10
|
cannam@18
|
11 #include <vector>
|
cannam@18
|
12
|
cannam@18
|
13 #include "segment.h"
|
cannam@18
|
14 #include "Segmenter.h"
|
cannam@20
|
15 #include "hmm/hmm.h"
|
cannam@20
|
16 #include "base/Window.h"
|
cannam@18
|
17
|
cannam@18
|
18 using std::vector;
|
cannam@18
|
19
|
cannam@24
|
20 class Decimator;
|
cannam@24
|
21 class ConstantQ;
|
cannam@26
|
22 class MFCC;
|
cannam@64
|
23 class FFTReal;
|
cannam@24
|
24
|
cannam@24
|
25 class ClusterMeltSegmenterParams
|
cannam@24
|
26 // defaults are sensible for 11025Hz with 0.2 second hopsize
|
cannam@18
|
27 {
|
cannam@18
|
28 public:
|
cannam@24
|
29 ClusterMeltSegmenterParams() :
|
cannam@24
|
30 featureType(FEATURE_TYPE_CONSTQ),
|
cannam@24
|
31 hopSize(0.2),
|
cannam@24
|
32 windowSize(0.6),
|
cannam@24
|
33 fmin(62),
|
cannam@24
|
34 fmax(16000),
|
cannam@24
|
35 nbins(8),
|
cannam@24
|
36 ncomponents(20),
|
cannam@24
|
37 nHMMStates(40),
|
cannam@24
|
38 nclusters(10),
|
cannam@24
|
39 histogramLength(15),
|
cannam@24
|
40 neighbourhoodLimit(20) { }
|
cannam@24
|
41 feature_types featureType;
|
cannam@24
|
42 double hopSize; // in secs
|
cannam@24
|
43 double windowSize; // in secs
|
cannam@24
|
44 int fmin;
|
cannam@24
|
45 int fmax;
|
cannam@24
|
46 int nbins;
|
cannam@24
|
47 int ncomponents;
|
cannam@24
|
48 int nHMMStates;
|
cannam@24
|
49 int nclusters;
|
cannam@24
|
50 int histogramLength;
|
cannam@24
|
51 int neighbourhoodLimit;
|
cannam@18
|
52 };
|
cannam@18
|
53
|
cannam@18
|
54 class ClusterMeltSegmenter : public Segmenter
|
cannam@18
|
55 {
|
cannam@18
|
56 public:
|
cannam@24
|
57 ClusterMeltSegmenter(ClusterMeltSegmenterParams params);
|
cannam@24
|
58 virtual ~ClusterMeltSegmenter();
|
cannam@24
|
59 virtual void initialise(int samplerate);
|
cannam@24
|
60 virtual int getWindowsize();
|
cannam@24
|
61 virtual int getHopsize();
|
cannam@24
|
62 virtual void extractFeatures(const double* samples, int nsamples);
|
cannam@24
|
63 void setFeatures(const vector<vector<double> >& f); // provide the features yourself
|
cannam@24
|
64 virtual void segment(); // segment into default number of segment-types
|
cannam@24
|
65 void segment(int m); // segment into m segment-types
|
cannam@24
|
66 int getNSegmentTypes() { return nclusters; }
|
cannam@24
|
67
|
cannam@18
|
68 protected:
|
cannam@24
|
69 void makeSegmentation(int* q, int len);
|
cannam@18
|
70
|
cannam@26
|
71 void extractFeaturesConstQ(const double *, int);
|
cannam@26
|
72 void extractFeaturesMFCC(const double *, int);
|
cannam@26
|
73
|
cannam@24
|
74 Window<double> *window;
|
cannam@64
|
75 FFTReal *fft;
|
cannam@26
|
76 ConstantQ* constq;
|
cannam@26
|
77 MFCC* mfcc;
|
cannam@24
|
78 model_t* model; // the HMM
|
cannam@24
|
79 int* q; // the decoded HMM state sequence
|
cannam@24
|
80 vector<vector<double> > histograms;
|
cannam@24
|
81
|
cannam@24
|
82 feature_types featureType;
|
cannam@24
|
83 double hopSize; // in seconds
|
cannam@24
|
84 double windowSize; // in seconds
|
cannam@24
|
85
|
cannam@24
|
86 // constant-Q parameters
|
cannam@24
|
87 int fmin;
|
cannam@24
|
88 int fmax;
|
cannam@24
|
89 int nbins;
|
cannam@24
|
90 int ncoeff;
|
cannam@24
|
91
|
cannam@24
|
92 // PCA parameters
|
cannam@24
|
93 int ncomponents;
|
cannam@24
|
94
|
cannam@24
|
95 // HMM parameters
|
cannam@24
|
96 int nHMMStates;
|
cannam@24
|
97
|
cannam@24
|
98 // clustering parameters
|
cannam@24
|
99 int nclusters;
|
cannam@24
|
100 int histogramLength;
|
cannam@24
|
101 int neighbourhoodLimit;
|
cannam@24
|
102
|
cannam@24
|
103 Decimator *decimator;
|
cannam@18
|
104 };
|