void CCorrelationFilters::prepare_data(struct CDataStruct *img, struct CParamStruct *params)
{
normalize_data(img);
img->size_data_freq = params->size_filt_freq;
fft_data(img);
params->num_elements_freq = img->num_elements_freq;
compute_psd_matrix(img, params);
compute_inverse_psd_matrix(img, params);
}
Patient(std::string name) {
this->name = name;
this->x = getData("x0l.dat");
this->y = getData("y0l.dat");
this->z = getData("z0l.dat");
(this->data).push_back(this->x);
(this->data).push_back(this->y);
(this->data).push_back(this->z);
normalize_data();
}
void similarity_objects(double *x, int *n, int *p, double *S, int *method)
{
/* similarity per variable */
void (*similarity)(double *, int , int , double *) = NULL;
switch(*method) {
case NUMERICAL_EUCLIDEAN:
Rprintf("LLAsim: the objects are assumed to be described by numerical variables\n");
similarity = similarity_numerical_euclidean;
break;
case NUMERICAL_COSINUS:
Rprintf("LLAsim: the objects are assumed to be described by numerical variables\n");
similarity = similarity_numerical_cosinus;
normalize_data(x, *n, *p);
break;
case CATEGORICAL:
Rprintf("LLAsim: the objects are assumed to be described by categorical variables\n");
similarity = similarity_categorical;
break;
case ORDINAL:
Rprintf("LLAsim: the objects are assumed to be described by ranks induced by ordinal variables\n");
similarity = similarity_ordinal;
break;
case BOOLEAN:
Rprintf("LLAsim: the objects are assumed to be described by boolean variables\n");
normalize_data(x, *n, *p);
similarity = similarity_numerical_cosinus;
break;
default:
error("invalid similarity method");
}
/* construction of the similarity by additive decomposition */
similarity(x,*n,*p, S);
/* normalization of the similarity index */
normalize_similarity(S,(*n) * (*n - 1)/2);
}
jdoubleArray Java_edu_cornell_audioProbe_AudioManager_features(JNIEnv* env, jobject javaThis, jshortArray array) {
//void Java_edu_cornell_audioProbe_AudioManager_features(JNIEnv* env, jobject javaThis, jshortArray array) {
(*env)->GetShortArrayRegion(env, array, 0, FRAME_LENGTH, buf);
normalize_data();
// apply window
computeHamming();
// computeFwdFFT
kiss_fftr(cfgFwd, normalizedData, fftx);
//compute power spectrum
computePowerSpec(fftx, powerSpec, FFT_LENGTH);
//compute magnitude spectrum
computeMagnitudeSpec(powerSpec, magnitudeSpec, FFT_LENGTH);
// compute total energy
energy = computeEnergy(powerSpec, FFT_LENGTH) / FFT_LENGTH;
//compute Spectral Entropy
computeSpectralEntropy2(magnitudeSpec, FFT_LENGTH);
//compute auto-correlation peaks
computeAutoCorrelationPeaks2(powerSpec, powerSpecCpx, NOISE_LEVEL, FFT_LENGTH);
//data output
////return data as variable size array caused by variable autocorrelation information.
jdoubleArray featureVector = (*env)->NewDoubleArray(env,6 + 2*numAcorrPeaks + 2 + LOOK_BACK_LENGTH);
//jdoubleArray featureVector = (*env)->NewDoubleArray(env,6 + 2*numAcorrPeaks + 2 + LOOK_BACK_LENGTH + FFT_LENGTH);
featuresValuesTemp[0] = numAcorrPeaks; //autocorrelation values
featuresValuesTemp[1] = maxAcorrPeakVal;
featuresValuesTemp[2] = maxAcorrPeakLag;
featuresValuesTemp[3] = spectral_entropy;
featuresValuesTemp[4] = rel_spectral_entropy;
featuresValuesTemp[5] = energy;
//gaussian distribution
//test the gaussian distribution with some dummy values first
x[0] = maxAcorrPeakVal;
x[1] = numAcorrPeaks;
x[2] = rel_spectral_entropy;
/*
emissionVoiced = computeMvnPdf(x,mean_voiced, inv_cov_voiced, denom_gauss_voiced);
emissionUnvoiced = computeMvnPdf(x,mean_unvoiced, inv_cov_unvoiced, denom_gauss_unvoiced);
*/
inferenceResult = getViterbiInference(x,featureAndInference);
memcpy( featuresValuesTemp+6, featureAndInference, (2+LOOK_BACK_LENGTH)*sizeof(double) ); //observation probabilities, inferences
//put auto correlation values in the string
memcpy( featuresValuesTemp+6+2+LOOK_BACK_LENGTH, acorrPeakValueArray, numAcorrPeaks*sizeof(double) );
memcpy( featuresValuesTemp+6+numAcorrPeaks+2+LOOK_BACK_LENGTH, acorrPeakLagValueArray, numAcorrPeaks*sizeof(double) );
//memcpy( featuresValuesTemp+6+numAcorrPeaks+numAcorrPeaks+2+LOOK_BACK_LENGTH, magnSpect, FFT_LENGTH*sizeof(double) );
(*env)->SetDoubleArrayRegion( env, featureVector, 0, 6 + numAcorrPeaks*2 + 2 + LOOK_BACK_LENGTH, (const jdouble*)featuresValuesTemp );
//(*env)->SetDoubleArrayRegion( env, featureVector, 0, 6 + numAcorrPeaks*2 + 2 + LOOK_BACK_LENGTH + FFT_LENGTH, (const jdouble*)featuresValuesTemp );
return featureVector;
}
static void
fit_do(FitControls *controls)
{
FitParamArg *arg;
FitArgs *args;
GtkWidget *dialog;
gdouble *param, *error;
gboolean *fixed;
gint i, j, nparams, nfree = 0;
gboolean allfixed, errorknown;
args = controls->args;
nparams = gwy_nlfit_preset_get_nparams(args->fitfunc);
fixed = g_newa(gboolean, nparams);
allfixed = TRUE;
for (i = 0; i < nparams; i++) {
arg = &g_array_index(args->param, FitParamArg, i);
fixed[i] = arg->fix;
allfixed &= fixed[i];
arg->value = arg->init;
if (!fixed[i])
nfree++;
}
if (allfixed)
return;
if (!normalize_data(args))
return;
if (gwy_data_line_get_res(args->xdata) <= nfree) {
dialog = gtk_message_dialog_new(GTK_WINDOW(controls->dialog),
GTK_DIALOG_DESTROY_WITH_PARENT,
GTK_MESSAGE_ERROR,
GTK_BUTTONS_OK,
_("It is necessary to select more "
"data points than free fit "
"parameters"));
gtk_dialog_run(GTK_DIALOG(dialog));
gtk_widget_destroy(dialog);
return;
}
if (args->fitter)
gwy_math_nlfit_free(args->fitter);
param = g_newa(gdouble, nparams);
error = g_newa(gdouble, nparams);
for (i = 0; i < nparams; i++)
param[i] =g_array_index(args->param, FitParamArg, i).value;
args->fitter
= gwy_nlfit_preset_fit(args->fitfunc, NULL,
gwy_data_line_get_res(args->xdata),
gwy_data_line_get_data_const(args->xdata),
gwy_data_line_get_data_const(args->ydata),
param, error, fixed);
errorknown = (args->fitter->covar != NULL);
for (i = 0; i < nparams; i++) {
arg = &g_array_index(args->param, FitParamArg, i);
arg->value = param[i];
arg->error = error[i];
fit_param_row_update_value(controls, i, errorknown);
}
if (errorknown) {
gchar buf[16];
/* FIXME: this is _scaled_ dispersion */
g_snprintf(buf, sizeof(buf), "%2.3g",
gwy_math_nlfit_get_dispersion(args->fitter));
gtk_label_set_markup(GTK_LABEL(controls->chisq), buf);
for (i = 0; i < nparams; i++) {
for (j = 0; j <= i; j++) {
g_snprintf(buf, sizeof(buf), "%0.3f",
gwy_math_nlfit_get_correlations(args->fitter, i, j));
fix_minus(buf, sizeof(buf));
gtk_label_set_markup(SLi(controls->covar, GtkLabel*, i, j),
buf);
}
}
}
else
