andrew@0: /*
andrew@0:  *  DynamicDynamicBayesianArray.cpp
andrew@0:  *  midiCannamReader
andrew@0:  *
andrew@0:  *  Created by Andrew on 17/07/2011.
andrew@0:  *  Copyright 2011 QMUL. All rights reserved.
andrew@0:  *
andrew@0:  */
andrew@0: 
andrew@0: #include "DynamicBayesianArray.h"
andrew@0: #include "math.h"
andrew@0: #include "ofMain.h"
andrew@0: 
andrew@0: DynamicBayesianArray::DynamicBayesianArray(){
andrew@0: 	
andrew@0: //	prior.createVector(240);
andrew@0: //	likelihood.createVector(240);
andrew@0: //	posterior.createVector(240);	
andrew@0: 	testVector.createVector(240);
andrew@0: 	testVector.addGaussianShape(100,10, 0.1);
andrew@0: 	
andrew@0: 	
andrew@0: 	likelihoodNoise = 0.5;
andrew@0: 	likelihoodMean = ARRAY_SIZE/2;
andrew@0: 	likelihoodStdDev = ARRAY_SIZE / 12;
andrew@0: 	initialiseArray();
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::initialiseArray(){
andrew@0: 	
andrew@0: 	//maximumIndex = 12;//change this
andrew@0: 	setGaussianPrior(ARRAY_SIZE/2, ARRAY_SIZE/1);	
andrew@0: 	setGaussianLikelihood(ARRAY_SIZE/2, ARRAY_SIZE/1);//likelihoodMean, likelihoodStdDev);
andrew@0: 	
andrew@0: 	calculatePosterior();
andrew@0: 	renormalisePosterior();
andrew@0: 	posteriorDecayRate = 0.06;
andrew@0: 	
andrew@0: 	eighthNoteProportion = 0.35;//must be less than 0.5 to discriminate - was 0.4
andrew@0: 	earlySixteenthNoteProportion = 0;
andrew@0: 	lateSixteenthNoteProportion = 0;
andrew@0: 	decayNoiseAmount = 0.1;
andrew@0: 	decayNoiseStdDev = ARRAY_SIZE/24;
andrew@0: 	standardDeviation = likelihoodStdDev;
andrew@0: 	setDecayNoiseGaussian(ARRAY_SIZE/2, decayNoiseStdDev);
andrew@0: 	
andrew@0: 	setGaussianLikelihood(likelihoodMean, likelihoodStdDev);
andrew@0: }
andrew@0: 
andrew@0: 
andrew@0: void DynamicBayesianArray::setGaussianPrior(float mean, float StdDev){
andrew@0: 	int i;
andrew@0: 	for (i=0;i<ARRAY_SIZE;i++){
andrew@0: 		prior[i] = (1/(StdDev*sqrt(2*PI)))*exp(-1*(i-mean)*(i-mean)/(2*StdDev*StdDev));
andrew@0: 		//posterior[i] = prior[i];
andrew@0: 	}
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::setGaussianPosterior(float mean, float StdDev){
andrew@0: 	int i;
andrew@0: 	for (i=0;i<ARRAY_SIZE;i++){
andrew@0: 		posterior[i] = (1/(StdDev*sqrt(2*PI)))*exp(-1*(i-mean)*(i-mean)/(2*StdDev*StdDev));
andrew@0: 	}
andrew@0: }
andrew@0: 
andrew@0: 
andrew@0: void DynamicBayesianArray::setGaussianLikelihood(float mean, float StdDev){
andrew@0: 	if (mean >= 0 && mean <= ARRAY_SIZE){
andrew@0: 		int i;	float eighthDifference;
andrew@0: 		int eighthPosition = ((int)mean + ARRAY_SIZE/2)%ARRAY_SIZE;
andrew@0: 		float mainDifference;
andrew@0: 		float gaussianProportion = 1 - likelihoodNoise;
andrew@0: 		
andrew@0: 		for (i=0;i < ARRAY_SIZE;i++){
andrew@0: 			
andrew@0: 			mainDifference = min( double(fabs(i-mean)) , (double)(i + ARRAY_SIZE - mean));
andrew@0: 			//without * (1 - eighthNoteProportion)
andrew@0: 			likelihood[i] = gaussianProportion  * (1/(StdDev*sqrt(2*PI)))*exp(-1*(mainDifference)*(mainDifference)/(2*StdDev*StdDev)) ;
andrew@0: 			
andrew@0: 			likelihood[i] += (likelihoodNoise / ARRAY_SIZE);
andrew@0: 			//likelihood[i] = (float) max(gaussianProportion * (1/(StdDev*sqrt(2*PI)))*exp(-1*(i-mean)*(i-mean)/(2*StdDev*StdDev)) , 
andrew@0: 			//(double) (likelihoodNoise / ARRAY_SIZE) );
andrew@0: 		}
andrew@0: 		//	renormaliseArray(&likelihood[0], ARRAY_SIZE);
andrew@0: 	}//end if mean within limits
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::calculatePosterior(){
andrew@0: 	int i;
andrew@0: 	for (i=0;i < ARRAY_SIZE;i++){
andrew@0: 		posterior[i] = likelihood[i] * prior[i];
andrew@0: 	}
andrew@0: 	//renormalisePosterior();
andrew@0: }
andrew@0: 
andrew@0: 
andrew@0: float DynamicBayesianArray::getMaximum(float *ptr, int length){
andrew@0: 	int i;
andrew@0: 	float max = 0;
andrew@0: 	for (i=0;i < length;i++){	
andrew@0: 		if (*(ptr+i)>max)
andrew@0: 			max = *(ptr+i);
andrew@0: 	}
andrew@0: 	maximumValue = max;
andrew@0: 	return max;
andrew@0: }
andrew@0: 
andrew@0: float* DynamicBayesianArray::getMaximumEstimate(float *ptr, int length){
andrew@0: 	float returnArray[2];
andrew@0: 	int i;
andrew@0: 	float max = 0;
andrew@0: 	maximumIndex = 0;
andrew@0: 	for (i=0;i < length;i++){	
andrew@0: 		if (*(ptr+i)>max){
andrew@0: 			max = *(ptr+i);
andrew@0: 			maximumIndex = i;
andrew@0: 		}
andrew@0: 	}
andrew@0: 	returnArray[0] = max;
andrew@0: 	returnArray[1] = maximumIndex;
andrew@0: 	maximumValue = max;
andrew@0: 	return &returnArray[0];
andrew@0: }
andrew@0: 
andrew@0: 
andrew@0: 
andrew@0: double DynamicBayesianArray::getIntegratedEstimateIndex(){
andrew@0: 	int i;
andrew@0: 	float integratedQuantity = 0;
andrew@0: 	float integratedTotal = 0;
andrew@0: 	double integratedIndex = 0;
andrew@0: 	for (i=0;i < ARRAY_SIZE;i++){	
andrew@0: 		integratedQuantity += posterior[i];//the values of the probability distribution
andrew@0: 		integratedTotal += i*posterior[i];
andrew@0: 	}
andrew@0: 	if (integratedQuantity > 0){
andrew@0: 		integratedIndex = integratedTotal / integratedQuantity;
andrew@0: 	}
andrew@0: 	integratedEstimate = (float) integratedIndex;
andrew@0: 	return integratedIndex;
andrew@0: }
andrew@0: 
andrew@0: 
andrew@0: double DynamicBayesianArray::calculateStandardDeviation(){
andrew@0: 	
andrew@0: 	double total = 0;
andrew@0: 	double pdfSum;
andrew@0: 	double variance = 0;
andrew@0: 	for (int i=0;i < ARRAY_SIZE;i++){	
andrew@0: 		//*posterior[i] * 		
andrew@0: 		total += posterior[i] * (i - integratedEstimate) * (i - integratedEstimate);//the values of the probability distribution
andrew@0: 		pdfSum += posterior[i];
andrew@0: 	}
andrew@0: 	
andrew@0: 	if (pdfSum > 0)
andrew@0: 		variance = total / pdfSum;
andrew@0: 	else 
andrew@0: 		variance = ARRAY_SIZE;
andrew@0: 	
andrew@0: 	standardDeviation = sqrt(variance);
andrew@0: 	return standardDeviation;
andrew@0: }
andrew@0: 
andrew@0: 
andrew@0: 
andrew@0: void DynamicBayesianArray::renormaliseArray(float *ptr, int length){
andrew@0: 	int i;
andrew@0: 	float totalArea = 0;
andrew@0: 	for (i=0;i < length;i++){
andrew@0: 		totalArea += *(ptr+i);
andrew@0: 	}
andrew@0: 	
andrew@0: 	for (i=0;i < length;i++){
andrew@0: 		*(ptr+i) /= totalArea;
andrew@0: 	}
andrew@0: 	
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::resetPrior(){
andrew@0: 	int i;
andrew@0: 	for (i=0;i<ARRAY_SIZE;i++){
andrew@0: 		prior[i] = posterior[i];
andrew@0: 	}
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::renormalisePosterior(){
andrew@0: 	int i;
andrew@0: 	float totalArea = 0;
andrew@0: 	for (i=0;i < ARRAY_SIZE;i++){
andrew@0: 		totalArea += posterior[i];
andrew@0: 	}
andrew@0: 	
andrew@0: 	for (i=0;i < ARRAY_SIZE;i++){
andrew@0: 		posterior[i] /= totalArea;
andrew@0: 	}
andrew@0: 	
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::decayPosterior(){
andrew@0: 	float *pointer;
andrew@0: 	pointer = getMaximumEstimate(&posterior[0], ARRAY_SIZE);	
andrew@0: 	float maximum;
andrew@0: 	maximum = *pointer;
andrew@0: 	int i;
andrew@0: 	for (i=0;i<ARRAY_SIZE;i++){
andrew@0: 		posterior[i] += (maximum - posterior[i]) * posteriorDecayRate * 0.01;;//usded to be * maximum not minus value
andrew@0: 	}
andrew@0: 	maximumIndex = *(pointer+1);
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::setDecayNoiseGaussian(float mean, float StdDev){
andrew@0: 	int i;
andrew@0: 	for (i=0;i<ARRAY_SIZE;i++){
andrew@0: 		decayNoiseArray[i] = (1/(StdDev*sqrt(2*PI)))*exp(-1*(i-mean)*(i-mean)/(2*StdDev*StdDev));
andrew@0: 	}
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::decayPosteriorWithGaussianNoise(){
andrew@0: 	
andrew@0: 	int i;
andrew@0: 	float currentMaximum = getMaximum(&posterior[0], ARRAY_SIZE);
andrew@0: 	for (i=0;i<ARRAY_SIZE;i++){
andrew@0: 		posterior[i] += decayNoiseArray[(i - (int)maximumIndex + ((3*ARRAY_SIZE)/2)) % ARRAY_SIZE] * currentMaximum * decayNoiseAmount;
andrew@0: 		//posteriorDecayRate * 0.01;;//usded to be * maximum not minus value
andrew@0: 	}
andrew@0: 	
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::resetMaximumPosterior(){
andrew@0: 	int i;
andrew@0: 	float max = 0;
andrew@0: 	for (i=0;i < ARRAY_SIZE;i++){	
andrew@0: 		if (posterior[i]>max){
andrew@0: 			maximumIndex = i;
andrew@0: 			max = posterior[i];
andrew@0: 		}
andrew@0: 	}
andrew@0: }
andrew@0: 
andrew@0: void DynamicBayesianArray::translateDistribution(int translationIndex){
andrew@0: 	int tmpIndex;
andrew@0: 	//copy array
andrew@0: 	int i;
andrew@0: 	for (i=0;i < ARRAY_SIZE;i++){
andrew@0: 		tempPosteriorArray[i] =	posterior[i] ;
andrew@0: 	}
andrew@0: 	//translate values
andrew@0: 	for (i=0;i < ARRAY_SIZE;i++){
andrew@0: 		tmpIndex = (i + translationIndex + ARRAY_SIZE)%ARRAY_SIZE;
andrew@0: 		posterior[tmpIndex] = tempPosteriorArray[i]; 
andrew@0: 	}
andrew@0: 	//now delete tmp array
andrew@0: }
andrew@0: 
andrew@0: 
andrew@0: void DynamicBayesianArray::drawFloatArray(float* arrayToDraw, const int& minIndex, const int& maxIndex){
andrew@0: 	
andrew@0: 	if (minIndex >= 0){
andrew@0: 	
andrew@0: 		double stepSize = ofGetWidth() / (double)(maxIndex - minIndex);
andrew@0: 		double screenHeight = ofGetHeight();
andrew@0: 		double maxVal = getMaximum(&arrayToDraw[0], maxIndex);
andrew@0: 
andrew@0: 		for (int i = minIndex+1;i < maxIndex;i++){
andrew@0: 		
andrew@0: 			ofLine (stepSize*(i-1), screenHeight * (1 - arrayToDraw[i-1] / maxVal), stepSize*i, screenHeight * (1 - arrayToDraw[i] / maxVal) );
andrew@0: 		}
andrew@0: 	
andrew@0: 	}
andrew@0: 		
andrew@0: 		
andrew@0: }
andrew@0: 
andrew@0:  
andrew@0: 
andrew@0: /*
andrew@0: void DynamicBayesianArray::drawDoubleArray(double[]& arrayToDraw, const int& minIndex, const int& maxIndex){
andrew@0: 
andrew@0: 	if (minIndex >= 0 && maxIndex <= arrayToDraw.size(0))
andrew@0: 
andrew@0: }
andrew@0: */
andrew@0: /*
andrew@0:  void DynamicBayesianArray::setGaussianLikelihoodForBeats(float mean, float StdDev){
andrew@0:  //this has eighth and sixteenth positions included
andrew@0:  
andrew@0:  if (mean >= 0 && mean <= ARRAY_SIZE){
andrew@0:  int i;	float eighthDifference;
andrew@0:  int eighthPosition = ((int)mean + ARRAY_SIZE/2)%ARRAY_SIZE;
andrew@0:  int earlySixteenthPosition = ((int)mean + (3*ARRAY_SIZE/4))%ARRAY_SIZE;;
andrew@0:  int lateSixteenthPosition = ((int)mean + (ARRAY_SIZE/4))%ARRAY_SIZE;;
andrew@0:  
andrew@0:  float mainDifference, sixteenthDifference;
andrew@0:  float gaussianProportion = 1 - likelihoodNoise;
andrew@0:  float mainProportion = (1 - eighthNoteProportion - earlySixteenthNoteProportion - lateSixteenthNoteProportion);
andrew@0:  
andrew@0:  for (i=0;i < ARRAY_SIZE;i++){
andrew@0:  
andrew@0:  mainDifference = min( fabs(i-mean) , (double)(i + ARRAY_SIZE - mean));
andrew@0:  likelihood[i] = gaussianProportion * mainProportion * (1/(StdDev*sqrt(2*PI)))*exp(-1*(mainDifference)*(mainDifference)/(2*StdDev*StdDev)) ;
andrew@0:  
andrew@0:  eighthDifference = min( abs(i - eighthPosition) , i + ARRAY_SIZE - eighthPosition);
andrew@0:  eighthDifference = min(eighthDifference , (float)(ARRAY_SIZE + eighthPosition - i ));
andrew@0:  //for e.g. +0.43, or -0.47 we require the gaussian around the half note too
andrew@0:  likelihood[i] += gaussianProportion * eighthNoteProportion * (1/(StdDev*sqrt(2*PI)))*exp(-1*(eighthDifference)*(eighthDifference)/(2*StdDev*StdDev)) ;
andrew@0:  
andrew@0:  sixteenthDifference = min( abs(i - earlySixteenthPosition) , i + ARRAY_SIZE - earlySixteenthPosition);
andrew@0:  sixteenthDifference = min(sixteenthDifference , (float)(ARRAY_SIZE + earlySixteenthPosition - i ));
andrew@0:  //for e.g. +0.43, or -0.47 we require the gaussian around the half note too
andrew@0:  likelihood[i] += gaussianProportion * earlySixteenthNoteProportion * (1/(StdDev*sqrt(2*PI)))*exp(-1*(sixteenthDifference)*(sixteenthDifference)/(2*StdDev*StdDev)) ;
andrew@0:  
andrew@0:  sixteenthDifference = min( abs(i - lateSixteenthPosition) , i + ARRAY_SIZE - lateSixteenthPosition);
andrew@0:  sixteenthDifference = min(sixteenthDifference , (float)(ARRAY_SIZE + lateSixteenthPosition - i ));
andrew@0:  //for e.g. +0.43, or -0.47 we require the gaussian around the half note too
andrew@0:  likelihood[i] += gaussianProportion * lateSixteenthNoteProportion * (1/(StdDev*sqrt(2*PI)))*exp(-1*(sixteenthDifference)*(sixteenthDifference)/(2*StdDev*StdDev)) ;
andrew@0:  
andrew@0:  
andrew@0:  
andrew@0:  likelihood[i] += (likelihoodNoise / ARRAY_SIZE);
andrew@0:  //likelihood[i] = (float) max(gaussianProportion * (1/(StdDev*sqrt(2*PI)))*exp(-1*(i-mean)*(i-mean)/(2*StdDev*StdDev)) , 
andrew@0:  //(double) (likelihoodNoise / ARRAY_SIZE) );
andrew@0:  }
andrew@0:  //	renormaliseArray(&likelihood[0], ARRAY_SIZE);
andrew@0:  }//end if mean within limits
andrew@0:  }
andrew@0:  */