annotate dsp/segmentation/SavedFeatureSegmenter.h @ 298:255e431ae3d4

* Key detector: when returning key strengths, use the peak value of the
  three underlying chromagram correlations (from 36-bin chromagram)
  corresponding to each key, instead of the mean.
  Rationale: This is the same method as used when returning the key value,
  and it's nice to have the same results in both returned value and plot.
  The peak performed better than the sum with a simple test set of triads,
  so it seems reasonable to change the plot to match the key output rather
  than the other way around.

* FFT: kiss_fftr returns only the non-conjugate bins, synthesise the rest
  rather than leaving them (perhaps dangerously) undefined.  Fixes an
  uninitialised data error in chromagram that could cause garbage results
  from key detector.

* Constant Q: remove precalculated values again, I reckon they're not
  proving such a good tradeoff.

/*
 *  SavedFeatureSegmenter.h
 *  soundbite
 *
 *  Created by Mark Levy on 23/03/2006.
 *  Copyright 2006 Centre for Digital Music, Queen Mary, University of London. All rights reserved.
 *
 */

#include <vector>

#include "segment.h"
#include "Segmenter.h"
#include "hmm/hmm.h"

using std::vector;

class SavedFeatureSegmenterParams		
{
public:
	SavedFeatureSegmenterParams() : hopSize(0.2), windowSize(0.6), 
	nHMMStates(40), nclusters(10), histogramLength(15), neighbourhoodLimit(20) { }
	double hopSize;		// in secs
	double windowSize;	// in secs
	int nHMMStates;
	int nclusters;
	int histogramLength;
	int neighbourhoodLimit;
};

class SavedFeatureSegmenter : public Segmenter
{
public:
	SavedFeatureSegmenter(SavedFeatureSegmenterParams params);
	virtual ~SavedFeatureSegmenter();
	virtual void initialise(int samplerate);
	virtual int getWindowsize() { return static_cast<int>(windowSize * samplerate); }
	virtual int getHopsize() { return static_cast<int>(hopSize * samplerate); }
	virtual void extractFeatures(double* samples, int nsamples) { }
	void setFeatures(const vector<vector<double> >& f);		// provide the features yourself
	virtual void segment();		// segment into default number of segment-types
	void segment(int m);		// segment into m segment-types
	int getNSegmentTypes() { return nclusters; }
protected:
	void makeSegmentation(int* q, int len);
	
	model_t* model;				// the HMM
	int* q;						// the decoded HMM state sequence
	vector<vector<double> > histograms;	
	
	double hopSize;		// in seconds
	double windowSize;	// in seconds
	
	// HMM parameters
	int nHMMStates;
	
	// clustering parameters
	int nclusters;
	int histogramLength;
	int neighbourhoodLimit;
};