//-----------------------------------------------------------------------------
// Test program.
// test.exe input.wav outout.wav f0 spec flag
// input.wav : argv[1] Input file
// output.wav : argv[2] Output file
// f0 : argv[3] F0 scaling (a positive number)
// spec : argv[4] Formant shift (a positive number)
//-----------------------------------------------------------------------------
int main(int argc, char *argv[]) {
if (argc != 2 && argc != 3 && argc != 4 && argc != 5) {
printf("error\n");
return 0;
}
int fs, nbit, x_length;
double *x = wavread(argv[1], &fs, &nbit, &x_length);
if (CheckLoadedFile(x, fs, nbit, x_length) == false) {
printf("error: File not found.\n");
return 0;
}
// Allocate memories
// The number of samples for F0
int f0_length = GetSamplesForDIO(fs, x_length, FRAMEPERIOD);
double *f0 = new double[f0_length];
double *time_axis = new double[f0_length];
// FFT size for CheapTrick
int fft_size = GetFFTSizeForCheapTrick(fs);
double **spectrogram = new double *[f0_length];
double **aperiodicity = new double *[f0_length];
for (int i = 0; i < f0_length; ++i) {
spectrogram[i] = new double[fft_size / 2 + 1];
aperiodicity[i] = new double[fft_size / 2 + 1];
}
// F0 estimation
F0Estimation(x, x_length, fs, f0_length, f0, time_axis);
// Spectral envelope estimation
SpectralEnvelopeEstimation(x, x_length, fs, time_axis, f0, f0_length,
spectrogram);
// Aperiodicity estimation by D4C
AperiodicityEstimation(x, x_length, fs, time_axis, f0, f0_length,
fft_size, aperiodicity);
// Note that F0 must not be changed until all parameters are estimated.
ParameterModification(argc, argv, fs, f0, f0_length, spectrogram);
// The length of the output waveform
int y_length =
static_cast<int>((f0_length - 1) * FRAMEPERIOD / 1000.0 * fs) + 1;
double *y = new double[y_length];
// Synthesis
WaveformSynthesis(f0, f0_length, spectrogram, aperiodicity, fft_size,
FRAMEPERIOD, fs, y_length, y);
// Output
wavwrite(y, y_length, fs, 16, argv[2]);
printf("complete.\n");
delete[] x;
delete[] time_axis;
delete[] f0;
delete[] y;
for (int i = 0; i < f0_length; ++i) {
delete[] spectrogram[i];
delete[] aperiodicity[i];
}
delete[] spectrogram;
delete[] aperiodicity;
return 0;
}
//-----------------------------------------------------------------------------
// Test program.
// test.exe input.wav outout.wav f0 spec flag
// input.wav : argv[1] Input file
// output.wav : argv[2] Output file
// f0 : argv[3] F0 scaling (a positive number)
// spec : argv[4] Formant shift (a positive number)
//-----------------------------------------------------------------------------
int main(int argc, char *argv[]) {
if (argc != 2 && argc != 3 && argc != 4 && argc != 5) {
printf("error\n");
return -2;
}
// 2016/01/28: Important modification.
// Memory allocation is carried out in advanse.
// This is for compatibility with C language.
int x_length = GetAudioLength(argv[1]);
if (x_length <= 0) {
if (x_length == 0)
printf("error: File not found.\n");
else
printf("error: The file is not .wav format.\n");
return -1;
}
double *x = new double[x_length];
// wavread() must be called after GetAudioLength().
int fs, nbit;
wavread(argv[1], &fs, &nbit, x);
DisplayInformation(fs, nbit, x_length);
//---------------------------------------------------------------------------
// Analysis part
//---------------------------------------------------------------------------
// 2016/02/02
// A new struct is introduced to implement safe program.
WorldParameters world_parameters = { 0 };
// You must set fs and frame_period before analysis/synthesis.
world_parameters.fs = fs;
// 5.0 ms is the default value.
// Generally, the inverse of the lowest F0 of speech is the best.
// However, the more elapsed time is required.
world_parameters.frame_period = 5.0;
// F0 estimation
// DIO
// F0EstimationDio(x, x_length, &world_parameters);
// Harvest
F0EstimationHarvest(x, x_length, &world_parameters);
// Spectral envelope estimation
SpectralEnvelopeEstimation(x, x_length, &world_parameters);
// Aperiodicity estimation by D4C
AperiodicityEstimation(x, x_length, &world_parameters);
// Note that F0 must not be changed until all parameters are estimated.
ParameterModification(argc, argv, fs, world_parameters.f0_length,
world_parameters.fft_size, world_parameters.f0,
world_parameters.spectrogram);
//---------------------------------------------------------------------------
// Synthesis part (2016/04/19)
// There are three samples in speech synthesis
// 1: Conventional synthesis
// 2: Example of real-time synthesis
// 3: Example of real-time synthesis (Ring buffer is efficiently used)
//---------------------------------------------------------------------------
char filename[100];
// The length of the output waveform
int y_length = static_cast<int>((world_parameters.f0_length - 1) *
world_parameters.frame_period / 1000.0 * fs) + 1;
double *y = new double[y_length];
// Synthesis 1 (conventional synthesis)
for (int i = 0; i < y_length; ++i) y[i] = 0.0;
WaveformSynthesis(&world_parameters, fs, y_length, y);
sprintf(filename, "01%s", argv[2]);
wavwrite(y, y_length, fs, 16, filename);
// Synthesis 2 (All frames are added at the same time)
for (int i = 0; i < y_length; ++i) y[i] = 0.0;
WaveformSynthesis2(&world_parameters, fs, y_length, y);
sprintf(filename, "02%s", argv[2]);
wavwrite(y, y_length, fs, 16, filename);
// Synthesis 3 (Ring buffer is efficiently used.)
for (int i = 0; i < y_length; ++i) y[i] = 0.0;
WaveformSynthesis3(&world_parameters, fs, y_length, y);
sprintf(filename, "03%s", argv[2]);
wavwrite(y, y_length, fs, 16, filename);
delete[] y;
delete[] x;
DestroyMemory(&world_parameters);
printf("complete.\n");
return 0;
}
int main (int argc, char **argv)
{
float mfcc_result[NUMFILTERBANK-1];
FILE *fptr;
float Min=0.0f,Max=0.0f;
int16_t *samples = NULL;
uint16_t numFrame = 0;
uint16_t i,j,k;
int32_t Val_RGB;
uint32_t idxCoeff;
int tmp;
char Len;
Len = strlen(argv[1]);
for ( i=0; i < Len-4; i++)
{
FileOut[i]= argv[1][i];
FileOutTxt[i]= argv[1][i];
}
FileOut[Len-4] = '.';
FileOut[Len-3] = 'b';
FileOut[Len-2] = 'm';
FileOut[Len-1] = 'p';
FileOut[Len] = '\0';
FileOutTxt[Len-4] = '.';
FileOutTxt[Len-3] = 't';
FileOutTxt[Len-2] = 'x';
FileOutTxt[Len-1] = 't';
FileOutTxt[Len] = '\0';
/* Sotarage MFCC coeffience */
fptr=fopen(FileOutTxt,"w");
PreCalcFilterBank(FilterBank,fNorm, NUMBINHALF, NUMFILTERBANK);
//printf("Length: %d \n", strlen(argv[1]));
wavread(argv[1], &samples);
printf("No. of channels: %d\n", header->num_channels);
printf("Sample rate: %d\n", header->sample_rate);
printf("Bit rate: %dkbps\n", header->byte_rate*8 / 1000);
printf("Bits per sample: %d\n\n", header->bps);
//printf("Sample 0: %d\n", samples[0]);
//printf("Sample 1: %d\n", samples[1]);
// Modify the header values & samples before writing the new file
wavwrite("track2.wav", samples);
numFrame = header->datachunk_size/(2*2*NUMBINHALF) ;
//printf("Num Frame: %d \n", numFrame );
/*
BMP = (RGB_data **)malloc((NUMFILTERBANK-1)*sizeof(RGB_data*));
for (i=0; i < NUMFILTERBANK-1; i++)
{
BMP[i] = (RGB_data *)malloc(sizeof(RGB_data)*(2*numFrame)) ;
if(BMP[i] == NULL)
{
fprintf(stderr, "out of memory\n");
exit (0);
}
}
*/
BMP = (RGB_data *)malloc((NUMFILTERBANK-1)*sizeof(RGB_data*)*2*numFrame);
//memset(BMP, 0, sizeof(BMP));
for (i = 0; i < 2*numFrame-1; i++)
{
for (j=0; j< 2*NUMBINHALF;j++)
{
V[j].Re = (float)(samples[i*NUMBINHALF + j]);
//printf(" %d", samples[i*NUMBINHALF + j]);
V[j].Im = (float)0.0f;
}
//printf("\n");
MFCC(V, FilterBank ,fNorm, mfcc_result);
//printf("MFCC:");
for(idxCoeff = 0; idxCoeff < NUMFILTERBANK -1; idxCoeff++)
{
fprintf(fptr,"%f ", mfcc_result[idxCoeff]);
//Val_RGB = (int)(255/(63.164356+46.877232)*(mfcc_result[idxCoeff]+10));
Val_RGB = (int)(255*(mfcc_result[idxCoeff]));
//Val_RGB = tmp*tmp;
//Val_RGB = Val_RGB/255;
if (Val_RGB < 0) Val_RGB = 0;
if (Val_RGB > 255) Val_RGB = 255;
BMP[(NUMFILTERBANK - idxCoeff-2)*2*numFrame +i].g = (BYTE)(Val_RGB);
BMP[(NUMFILTERBANK - idxCoeff-2)*2*numFrame +i].b = (BYTE)(Val_RGB);
BMP[(NUMFILTERBANK - idxCoeff-2)*2*numFrame +i].r = (BYTE)(Val_RGB);
//BMP[NUMFILTERBANK - idxCoeff-1][i].b = (BYTE)((Val_RGB & 0x00FF00)>>8);
//BMP[NUMFILTERBANK - idxCoeff-1][i].r = (BYTE)((Val_RGB & 0xFF0000)>>16);
//printf(" %d ", Val_RGB);
//if (mfcc_result[idxCoeff] < Min) Min=mfcc_result[idxCoeff];
//if (mfcc_result[idxCoeff] > Max) Max=mfcc_result[idxCoeff];
}
fprintf(fptr,"\r\n");
}
//printf("Min: %f Max: %f \n", Min, Max);
bmp_generator(FileOut, 2*numFrame, NUMFILTERBANK -1 ,(BYTE*) (BMP));
//bmpread("record_21.bmp");
free(header);
free(samples);
free(BMP);
fclose(fptr);
exit (0);
}
int main()
{
//=============================================================================
//STEP 1 - INITIALISATION
//=============================================================================
//Create input and output SIGNAL structures.
SIGNAL input,output;
int i=0;
//Initialise SIGNAL structures.
memset(&input,0,sizeof(SIGNAL));
memset(&output,0,sizeof(SIGNAL));
//=============================================================================
//STEP 2 - OPEN AUDIO FILE AND READ INTO MEMORY
//=============================================================================
//Read input signal into memory.
printf("Reading %s into memory...\n",kInputSignalLocation);
wavread(kInputSignalLocation, &input);
//=============================================================================
//STEP 3 - PREPARE THE OUTPUT SIGNAL
//=============================================================================
//Set up output SIGNAL structure.
Fill_SIGNAL(&output,
input.frames,
input.samplerate,
input.channels,
SF_FORMAT_WAV|SF_FORMAT_PCM_32,
input.sections,
input.seekable,
NULL);
//Allocate space for the 'output.data' buffer.
output.data=(float *)malloc(sizeof(float)* input.frames*input.channels);
//=============================================================================
//STEP 4 - PROCESSING
//=============================================================================
//In this example we reverse the input signal
//and store the results in the output signal
for(i=0;i<input.frames;i++){
output.data[i]=input.data[input.frames-i];
//If the signal is stereo then a frame comprises two samples
if (input.channels==2)
output.data[i+1]=input.data[input.frames-i+1];
}
//=============================================================================
//STEP 5 - WRITE OUTPUT TO AUDIO FILE
//=============================================================================
//Write output file to disk
printf("\n\nWriting audio to output file %s\n",kOutputSignalLocation);
wavwrite(kOutputSignalLocation, &output);
//=============================================================================
//STEP 6 - CLEAN UP
//=============================================================================
free(input.data);
free(output.data);
return 0;
}
//-----------------------------------------------------------------------------
// Test program.
// test.exe input.wav outout.wav f0 spec flag
// input.wav : argv[1] Input file
// output.wav : argv[2] Output file
// f0 : argv[3] F0 scaling (a positive number)
// spec : argv[4] Formant shift (a positive number)
//-----------------------------------------------------------------------------
int main(int argc, char *argv[]) {
if (argc != 2 && argc != 3 && argc != 4 && argc != 5) {
printf("error\n");
return -2;
}
// 2016/01/28: Important modification.
// Memory allocation is carried out in advanse.
// This is for compatibility with C language.
int x_length = GetAudioLength(argv[1]);
if (x_length <= 0) {
if (x_length == 0)
printf("error: File not found.\n");
else
printf("error: The file is not .wav format.\n");
return -1;
}
double *x = (double*) malloc(sizeof(double) * (x_length));
// wavread() must be called after GetAudioLength().
int fs, nbit;
wavread(argv[1], &fs, &nbit, x);
DisplayInformation(fs, nbit, x_length);
//---------------------------------------------------------------------------
// Analysis part
//---------------------------------------------------------------------------
// 2016/02/02
// A new struct is introduced to implement safe program.
WorldParameters world_parameters = { 0 };
// You must set fs and frame_period before analysis/synthesis.
world_parameters.fs = fs;
// 5.0 ms is the default value.
// Generally, the inverse of the lowest F0 of speech is the best.
// However, the more elapsed time is required.
world_parameters.frame_period = 5.0;
// F0 estimation
F0Estimation(x, x_length, &world_parameters);
// Spectral envelope estimation
SpectralEnvelopeEstimation(x, x_length, &world_parameters);
// Aperiodicity estimation by D4C
AperiodicityEstimation(x, x_length, &world_parameters);
// Note that F0 must not be changed until all parameters are estimated.
ParameterModification(argc, argv, fs, world_parameters.f0_length,
world_parameters.fft_size, world_parameters.f0,
world_parameters.spectrogram);
//---------------------------------------------------------------------------
// Synthesis part
//---------------------------------------------------------------------------
// The length of the output waveform
int y_length = (int)((world_parameters.f0_length - 1) *
world_parameters.frame_period / 1000.0 * fs) + 1;
double *y = (double*) malloc(sizeof(double) * (y_length));
// Synthesis
WaveformSynthesis(&world_parameters, fs, y_length, y);
// Output
wavwrite(y, y_length, fs, 16, argv[2]);
if(y != NULL) {
free(y);
y = NULL;
}
if(x != NULL) {
free(x);
x = NULL;
}
DestroyMemory(&world_parameters);
printf("complete.\n");
return 0;
}
void wavwriteclose( WAV *wav )
{
if ( wav->nsamples )
wavwrite( wav, wav->nsamples, wav->samples[0], wav->samples[1] );
wavclose( wav );
}
/**
* Main function
*
*/
int main(int argc, char *argv[])
{
if (argc != 5)
{
fprintf(stderr, "%s <input_f0_file> <input_spectrum_file> <input_aperiodicity_file> <output_wav_file>", argv[0]);
return EXIT_FAILURE;
}
// Define a default filled structures
WorldParameters world_parameters;
world_parameters.fs = 22050; // FIXME: hardcoded value
world_parameters.f0_length = filesize(argv[1]) / sizeof(double);
world_parameters.fft_size = ((filesize(argv[2]) / (sizeof(double) * world_parameters.f0_length)) - 1) * 2; // Be careful that .sp contains only first half of the spectrum
std::cout << "fft size = " << world_parameters.fft_size << std::endl;
// 5.0 ms is the default value.
// Generally, the inverse of the lowest F0 of speech is the best.
// However, the more elapsed time is required.
world_parameters.frame_period = DEFAULT_FRAME_PERIOD;
//---------------------------------------------------------------------------
// Prepare memory
//---------------------------------------------------------------------------
world_parameters.f0 = new double[world_parameters.f0_length];
world_parameters.spectrogram = new double*[world_parameters.f0_length];
for (int i=0;i<world_parameters.f0_length; i++)
world_parameters.spectrogram[i] = new double[world_parameters.fft_size / 2 + 1];
world_parameters.aperiodicity = new double*[world_parameters.f0_length];
for (int i=0;i<world_parameters.f0_length; i++)
world_parameters.aperiodicity[i] = new double[world_parameters.fft_size / 2 + 1];
//---------------------------------------------------------------------------
// Loading
//---------------------------------------------------------------------------
// F0 loading
std::ifstream is_f0(argv[1], std::ios::binary | std::ios::in);
if ( !is_f0.is_open() )
return false;
is_f0.read(reinterpret_cast<char*>(world_parameters.f0),
std::streamsize(world_parameters.f0_length*sizeof(double)));
// for (int i=0; i<world_parameters.f0_length; i++)
// std::cout << world_parameters.f0[i] << std::endl;
is_f0.close();
// Spectrogram loading
std::ifstream is_spectrogram(argv[2], std::ios::binary | std::ios::in);
if ( !is_spectrogram.is_open() )
return false;
for (int i=0; i<world_parameters.f0_length; i++)
{
is_spectrogram.read(reinterpret_cast<char*>(world_parameters.spectrogram[i]),
std::streamsize((world_parameters.fft_size / 2 + 1)*sizeof(double)));
}
is_spectrogram.close();
// Aperiodicity loading
std::ifstream is_aperiodicity(argv[3], std::ios::binary | std::ios::in);
if ( !is_aperiodicity.is_open() )
return false;
for (int i=0; i<world_parameters.f0_length; i++)
{
is_aperiodicity.read(reinterpret_cast<char*>(world_parameters.aperiodicity[i]),
std::streamsize((world_parameters.fft_size / 2 + 1)*sizeof(double)));
}
is_aperiodicity.close();
//---------------------------------------------------------------------------
// Synthesis
//---------------------------------------------------------------------------
int y_length = static_cast<int>((world_parameters.f0_length - 1) *
world_parameters.frame_period / 1000.0 * world_parameters.fs) + 1;
double *y = new double[y_length];
for (int i = 0; i < y_length; ++i) y[i] = 0.0;
WaveformSynthesis(&world_parameters, y);
wavwrite(y, y_length, world_parameters.fs, 16, argv[4]);
//---------------------------------------------------------------------------
// Cleaning part
//---------------------------------------------------------------------------
delete[] y;
DestroyMemory(&world_parameters);
std::cout << "complete" << std::endl;
return EXIT_SUCCESS;
}
//-----------------------------------------------------------------------------
// Test program.
// test.exe input.wav outout.wav f0 spec flag
// input.wav : argv[1] Input file
// output.wav : argv[2] Output file
// f0 : argv[3] F0 scaling (a positive number)
// spec : argv[4] Formant shift (a positive number)
//-----------------------------------------------------------------------------
int main(int argc, char *argv[]) {
if (argc != 7) {
printf("command: synth FFT_length sampling_rate F0_file spectrogram_file aperiodicity_file output_waveform\n");
return -2;
}
int fft_size = atoi(argv[1]);
int fs = atoi(argv[2]);
// compute n bands from fs as in d4c.cpp:325
int number_of_aperiodicities = static_cast<int>(MyMinDouble(world::kUpperLimit, fs / 2.0 -
world::kFrequencyInterval) / world::kFrequencyInterval);
WorldParameters world_parameters = { 0 };
// You must set fs and frame_period before analysis/synthesis.
world_parameters.fs = fs;
// 5.0 ms is the default value.
// Generally, the inverse of the lowest F0 of speech is the best.
// However, the more elapsed time is required.
world_parameters.frame_period = 5.0;
world_parameters.fft_size = fft_size;
// find number of frames (doubles) in f0 file:
struct stat st;
if (stat(argv[3], &st) == -1) {
printf("cannot read f0\n");
return -2;
}
int f0_length = (st.st_size / sizeof(double));
world_parameters.f0_length = f0_length;
// printf("%d\n", f0_length);
world_parameters.f0 = new double[f0_length];
FILE *fp;
fp = fopen(argv[3], "rb");
for (int i = 0; i < f0_length; i++) {
fread(&world_parameters.f0[i], sizeof(double), 1, fp);
}
fclose(fp);
double **coarse_aperiodicities = new double *[world_parameters.f0_length];
world_parameters.aperiodicity = new double *[world_parameters.f0_length];
for (int i = 0; i < world_parameters.f0_length; ++i) {
world_parameters.aperiodicity[i] = new double[fft_size / 2 + 1];
coarse_aperiodicities[i] = new double[number_of_aperiodicities];
}
world_parameters.spectrogram = new double *[world_parameters.f0_length];
for (int i = 0; i < world_parameters.f0_length; ++i) {
world_parameters.spectrogram[i] = new double[fft_size / 2 + 1];
}
fp = fopen(argv[4], "rb");
for (int i = 0; i < f0_length; i++) {
for (int j = 0; j < fft_size / 2 + 1; j++) {
fread(&world_parameters.spectrogram[i][j], sizeof(double), 1, fp);
}
}
fclose(fp);
// aper
fp = fopen(argv[5], "rb");
for (int i = 0; i < f0_length; i++) {
for (int j = 0; j < number_of_aperiodicities; j++) {
fread(&coarse_aperiodicities[i][j], sizeof(double), 1, fp);
}
}
fclose(fp);
// convert bandaps to full aperiodic spectrum by interpolation (originally in d4c extraction):
// Linear interpolation to convert the coarse aperiodicity into its
// spectral representation.
// -- for interpolating --
double *coarse_aperiodicity = new double[number_of_aperiodicities + 2];
coarse_aperiodicity[0] = -60.0;
coarse_aperiodicity[number_of_aperiodicities + 1] = 0.0;
double *coarse_frequency_axis = new double[number_of_aperiodicities + 2];
for (int i = 0; i <= number_of_aperiodicities; ++i)
coarse_frequency_axis[i] =
static_cast<double>(i) * world::kFrequencyInterval;
coarse_frequency_axis[number_of_aperiodicities + 1] = fs / 2.0;
double *frequency_axis = new double[fft_size / 2 + 1];
for (int i = 0; i <= fft_size / 2; ++i)
frequency_axis[i] = static_cast<double>(i) * fs / fft_size;
// ----
for (int i = 0; i < f0_length; ++i) {
// load band ap values for this frame into coarse_aperiodicity
for (int k = 0; k < number_of_aperiodicities; ++k) {
coarse_aperiodicity[k+1] = coarse_aperiodicities[i][k];
}
interp1(coarse_frequency_axis, coarse_aperiodicity,
number_of_aperiodicities + 2, frequency_axis, fft_size / 2 + 1,
world_parameters.aperiodicity[i]);
for (int j = 0; j <= fft_size / 2; ++j)
world_parameters.aperiodicity[i][j] = pow(10.0, world_parameters.aperiodicity[i][j] / 20.0);
}
//printf("%d %d\n", world_parameters.f0_length, fs);
//---------------------------------------------------------------------------
// Synthesis part
//---------------------------------------------------------------------------
// The length of the output waveform
int y_length = static_cast<int>((world_parameters.f0_length - 1) *
FRAMEPERIOD / 1000.0 * fs) + 1;
double *y = new double[y_length];
// Synthesis
WaveformSynthesis(&world_parameters, fs, y_length, y);
// Output
wavwrite(y, y_length, fs, 16, argv[6]);
delete[] y;
DestroyMemory(&world_parameters);
for (int i=0; i<f0_length; i++){
delete[] coarse_aperiodicities[i];
}
delete[] coarse_aperiodicities;
delete[] coarse_aperiodicity;
delete[] frequency_axis;
printf("complete %s.\n", argv[6]);
return 0;
}
