Форум на Исходниках.RU

Скрытый текст

#include "aquila/global.h"

#include "aquila/source/WaveFile.h"

#include "aquila/source/SignalSource.h"

#include "aquila/transform/Mfcc.h"

#include "aquila/ml/Dtw.h"

#include "aquila/ml/DtwPoint.h"

#include "aquila/transform/FftFactory.h"

#include <iostream>

#include <vector>

using namespace std;

/* Rounding number of samples */

unsigned int RoundSize (Aquila::WaveFile& wav, int& Sample)

{

int SIZE = 0;

if (wav.getSamplesCount()%Sample == 0)

{

SIZE = wav.getSamplesCount();

}

else

{

SIZE = wav.getSamplesCount()-(wav.getSamplesCount()%Sample);

}

return SIZE;

}

/* Transform wavfile to frame matrix with double overlap */

vector< vector<double> > MatrixFromFile (Aquila::WaveFile& wav, int Row, int Col,int Window)

{

/* Filling the matrix */

vector< vector<double> > WindowMatrix(Row);

int PairCount = 0;

for(int i = 0; i < Row; i=i+2)

{

WindowMatrix[i].resize(Col);

for(int j = 0; j < Col; ++j)

{

WindowMatrix[i][j] = wav.sample(PairCount++);

}

int NoPairCount = Window/2;

for(int i = 1; i < Row; i=i+2)

{

WindowMatrix[i].resize(Col);

for(int j = 0; j < Col; j++)

{

WindowMatrix[i][j] = wav.sample(NoPairCount++);

}

return WindowMatrix;

}

/* Calculation of Mel Frequency Cepstral Coefficients for FrameMatrix with subtracting the average */

vector< vector<double> > CalcMFCCforMatrix (vector< vector<double> >& WindowMatrix, int FFTSize, int Row)

{

vector< vector<double> > result (Row);

for(int i = 0; i < Row; ++i)

{

result[i].resize(FFTSize);

Aquila::SignalSource WavFromArr(WindowMatrix.at(i),16000);

Aquila::Mfcc fftMFCC (FFTSize);

result.at(i) = fftMFCC.calculate(WavFromArr);

}

/* Searching the average */

vector<double> AverangeMFCC;

for(int i = 0; i < 12; ++i)

{

double TempSum = 0.0;

for(int j = 0; j < Row; ++j)

{

TempSum+= result[j][i];

}

AverangeMFCC.push_back(TempSum/Row);

}

/* Subtracting the average */

for(int i = 0; i < 12; ++i)

{

for(int j = 0; j < Row; ++j)

{

result[j][i] = (result[j][i]) - AverangeMFCC.at(i);

}

return result;

}

int main()

{

Aquila::WaveFile wav("d:\QT\Projects\console\build-wav-Desktop_Qt_5_0_2_MinGW_32bit-Debug\debug\Numbers\one.wav");

/* firs file */

int Window = 500; /*16kHz = 62.5uS per sample*/

unsigned int Wav1Size = RoundSize (wav,Window);

/* Matrix vector< vector<double> > size:

* Row - number of windows (one sample = 62.5uS, one WindowSize = sample*window, double overlap)

* Column - Window*/

const int RowWav1 = 2*(Wav1Size/Window);

const int ColWav1 = Window;

/*Wav file to FrameMatrix*/

vector< vector<double> > WindowMatrixWav1 = MatrixFromFile (wav,RowWav1,ColWav1,Window);

/* Calculation of Mel Frequency Cepstral Coefficients */

unsigned int FFTSize = 32;

vector< vector<double> > Wav1MatrixMFCC = CalcMFCCforMatrix (WindowMatrixWav1,FFTSize,RowWav1);

/* second file */

std::cout<<"--------"<<std::endl;

Aquila::WaveFile wav2("d:\QT\Projects\console\build-wav-Desktop_Qt_5_0_2_MinGW_32bit-Debug\debug\Numbers\four.wav");

unsigned int Wav2Size = RoundSize (wav2,Window);

/* Matrix vector< vector<double> > size:

* Row - number of windows (one sample = 62.5uS, one WindowSize = sample*window, double overlap)

* Column - Window*/

const int RowWav2 = 2*(Wav2Size/Window);

const int ColWav2 = Window;

/*Wav file to FrameMatrix*/

vector< vector<double> > WindowMatrixWav2 = MatrixFromFile (wav2,RowWav2,ColWav2,Window);

/* Calculation of Mel Frequency Cepstral Coefficients */

vector< vector<double> > Wav2MatrixMFCC = CalcMFCCforMatrix (WindowMatrixWav2,FFTSize,RowWav2);

/*DTW comparison*/

Aquila::DtwDataType from, to;

from = Wav1MatrixMFCC;

to = Wav2MatrixMFCC;

Aquila::Dtw dtw;

double distance = dtw.getDistance(from, to);

std::cout << "DTW distance (neighbors): " << distance << "\n";

Aquila::Dtw dtwDiag(Aquila::euclideanDistance, Aquila::Dtw::Diagonals);

distance = dtwDiag.getDistance(from, to);

std::cout << "DTW distance (diagonals): " << distance << "\n";

return 0;

}