cannam@18
|
1 /*
|
cannam@18
|
2 * ClusterMeltSegmenter.h
|
cannam@18
|
3 * soundbite
|
cannam@18
|
4 *
|
cannam@18
|
5 * Created by Mark Levy on 23/03/2006.
|
cannam@18
|
6 * Copyright 2006 Centre for Digital Music, Queen Mary, University of London. All rights reserved.
|
cannam@18
|
7 *
|
cannam@18
|
8 */
|
cannam@18
|
9
|
cannam@18
|
10 #include <vector>
|
cannam@18
|
11
|
cannam@18
|
12 #include "segment.h"
|
cannam@18
|
13 #include "Segmenter.h"
|
cannam@20
|
14 #include "hmm/hmm.h"
|
cannam@20
|
15 #include "base/Window.h"
|
cannam@20
|
16 #include "dsp/chromagram/ConstantQ.h"
|
cannam@18
|
17
|
cannam@18
|
18 using std::vector;
|
cannam@18
|
19
|
cannam@18
|
20 class ClusterMeltSegmenterParams // defaults are sensible for 11025Hz with 0.2 second hopsize
|
cannam@18
|
21 {
|
cannam@18
|
22 public:
|
cannam@18
|
23 ClusterMeltSegmenterParams() : featureType(FEATURE_TYPE_CONSTQ), hopSize(0.2), windowSize(0.6), fmin(62), fmax(16000),
|
cannam@18
|
24 nbins(8), ncomponents(20), nHMMStates(40), nclusters(10), histogramLength(15), neighbourhoodLimit(20) { }
|
cannam@18
|
25 feature_types featureType;
|
cannam@18
|
26 double hopSize; // in secs
|
cannam@18
|
27 double windowSize; // in secs
|
cannam@18
|
28 int fmin;
|
cannam@18
|
29 int fmax;
|
cannam@18
|
30 int nbins;
|
cannam@18
|
31 int ncomponents;
|
cannam@18
|
32 int nHMMStates;
|
cannam@18
|
33 int nclusters;
|
cannam@18
|
34 int histogramLength;
|
cannam@18
|
35 int neighbourhoodLimit;
|
cannam@18
|
36 };
|
cannam@18
|
37
|
cannam@18
|
38 class ClusterMeltSegmenter : public Segmenter
|
cannam@18
|
39 {
|
cannam@18
|
40 public:
|
cannam@18
|
41 ClusterMeltSegmenter(ClusterMeltSegmenterParams params);
|
cannam@18
|
42 virtual ~ClusterMeltSegmenter();
|
cannam@18
|
43 virtual void initialise(int samplerate);
|
cannam@20
|
44 virtual int getWindowsize();
|
cannam@20
|
45 virtual int getHopsize();
|
cannam@18
|
46 virtual void extractFeatures(double* samples, int nsamples);
|
cannam@18
|
47 void setFeatures(const vector<vector<double> >& f); // provide the features yourself
|
cannam@18
|
48 virtual void segment(); // segment into default number of segment-types
|
cannam@18
|
49 void segment(int m); // segment into m segment-types
|
cannam@18
|
50 int getNSegmentTypes() { return nclusters; }
|
cannam@18
|
51 protected:
|
cannam@18
|
52 //void mpeg7ConstQ();
|
cannam@18
|
53 void makeSegmentation(int* q, int len);
|
cannam@18
|
54
|
cannam@20
|
55 Window<double> *window;
|
cannam@20
|
56 // int windowLength; // in samples
|
cannam@20
|
57 ConstantQ* constq;
|
cannam@18
|
58 model_t* model; // the HMM
|
cannam@18
|
59 //vector<int> stateSequence;
|
cannam@18
|
60 //vector<int> segmentTypeSequence;
|
cannam@18
|
61 int* q; // the decoded HMM state sequence
|
cannam@18
|
62 vector<vector<double> > histograms;
|
cannam@18
|
63
|
cannam@18
|
64 feature_types featureType;
|
cannam@18
|
65 double hopSize; // in seconds
|
cannam@18
|
66 double windowSize; // in seconds
|
cannam@18
|
67
|
cannam@18
|
68 // constant-Q parameters
|
cannam@18
|
69 int fmin;
|
cannam@18
|
70 int fmax;
|
cannam@18
|
71 int nbins;
|
cannam@18
|
72 int ncoeff;
|
cannam@18
|
73
|
cannam@18
|
74 // PCA parameters
|
cannam@18
|
75 int ncomponents;
|
cannam@18
|
76
|
cannam@18
|
77 // HMM parameters
|
cannam@18
|
78 int nHMMStates;
|
cannam@18
|
79
|
cannam@18
|
80 // clustering parameters
|
cannam@18
|
81 int nclusters;
|
cannam@18
|
82 int histogramLength;
|
cannam@18
|
83 int neighbourhoodLimit;
|
cannam@18
|
84 };
|