void Manipulation_replaceOriginalSound (Manipulation me, Sound sound) { try { my sound = Sound_convertToMono (sound); Vector_subtractMean (my sound.get()); my lpc = autoLPC(); } catch (MelderError) { Melder_throw (me, U": original Sound not replaced with ", sound, U"."); } }
static LPC _Sound_to_LPC (Sound me, int predictionOrder, double analysisWidth, double dt, double preEmphasisFrequency, int method, double tol1, double tol2) { double t1, samplingFrequency = 1.0 / my dx; double windowDuration = 2 * analysisWidth; /* gaussian window */ long nFrames, frameErrorCount = 0; if (floor (windowDuration / my dx) < predictionOrder + 1) Melder_throw ("Analysis window duration too short.\n" "For a prediction order of ", predictionOrder, " the analysis window duration has to be greater than ", my dx * (predictionOrder + 1), "Please increase the analysis window duration or lower the prediction order."); // Convenience: analyse the whole sound into one LPC_frame if (windowDuration > my dx * my nx) { windowDuration = my dx * my nx; } Sampled_shortTermAnalysis (me, windowDuration, dt, & nFrames, & t1); autoSound sound = Data_copy (me); autoSound sframe = Sound_createSimple (1, windowDuration, samplingFrequency); autoSound window = Sound_createGaussian (windowDuration, samplingFrequency); autoLPC thee = LPC_create (my xmin, my xmax, nFrames, dt, t1, predictionOrder, my dx); autoMelderProgress progress (L"LPC analysis"); if (preEmphasisFrequency < samplingFrequency / 2) { Sound_preEmphasis (sound.peek(), preEmphasisFrequency); } for (long i = 1; i <= nFrames; i++) { LPC_Frame lpcframe = (LPC_Frame) & thy d_frames[i]; double t = Sampled_indexToX (thee.peek(), i); LPC_Frame_init (lpcframe, predictionOrder); Sound_into_Sound (sound.peek(), sframe.peek(), t - windowDuration / 2); Vector_subtractMean (sframe.peek()); Sounds_multiply (sframe.peek(), window.peek()); if (method == LPC_METHOD_AUTO) { if (! Sound_into_LPC_Frame_auto (sframe.peek(), lpcframe)) { frameErrorCount++; } } else if (method == LPC_METHOD_COVAR) { if (! Sound_into_LPC_Frame_covar (sframe.peek(), lpcframe)) { frameErrorCount++; } } else if (method == LPC_METHOD_BURG) { if (! Sound_into_LPC_Frame_burg (sframe.peek(), lpcframe)) { frameErrorCount++; } } else if (method == LPC_METHOD_MARPLE) { if (! Sound_into_LPC_Frame_marple (sframe.peek(), lpcframe, tol1, tol2)) { frameErrorCount++; } } if ( (i % 10) == 1) { Melder_progress ( (double) i / nFrames, L"LPC analysis of frame ", Melder_integer (i), L" out of ", Melder_integer (nFrames), L"."); } } return thee.transfer(); }
autoManipulation Sound_PointProcess_to_Manipulation (Sound sound, PointProcess point) { try { autoManipulation me = Manipulation_create (sound -> xmin, sound -> xmax); my sound = Sound_convertToMono (sound); Vector_subtractMean (my sound.get()); my pulses = Data_copy (point); my pitch = PointProcess_to_PitchTier (point, MAX_T); return me; } catch (MelderError) { Melder_throw (sound, U" & ", point, U": not converted to Manipulation."); } }
autoManipulation Sound_to_Manipulation (Sound me, double timeStep, double minimumPitch, double maximumPitch) { try { autoManipulation thee = Manipulation_create (my xmin, my xmax); thy sound = Sound_convertToMono (me); Vector_subtractMean (thy sound.get()); autoPitch pitch = Sound_to_Pitch (thy sound.get(), timeStep, minimumPitch, maximumPitch); thy pulses = Sound_Pitch_to_PointProcess_cc (thy sound.get(), pitch.get()); thy pitch = Pitch_to_PitchTier (pitch.get()); /* (DurationTier has been done at creation time) */ return thee; } catch (MelderError) { Melder_throw (me, U": not converted to Manipulation."); } }
PowerCepstrogram Sound_to_PowerCepstrogram (Sound me, double pitchFloor, double dt, double maximumFrequency, double preEmphasisFrequency) { try { // minimum analysis window has 3 periods of lowest pitch double analysisWidth = 3 / pitchFloor; double windowDuration = 2 * analysisWidth; /* gaussian window */ long nFrames; // Convenience: analyse the whole sound into one Cepstrogram_frame if (windowDuration > my dx * my nx) { windowDuration = my dx * my nx; } double t1, samplingFrequency = 2 * maximumFrequency; autoSound sound = Sound_resample (me, samplingFrequency, 50); Sound_preEmphasis (sound.peek(), preEmphasisFrequency); Sampled_shortTermAnalysis (me, windowDuration, dt, & nFrames, & t1); autoSound sframe = Sound_createSimple (1, windowDuration, samplingFrequency); autoSound window = Sound_createGaussian (windowDuration, samplingFrequency); // find out the size of the FFT long nfft = 2; while (nfft < sframe -> nx) nfft *= 2; long nq = nfft / 2 + 1; double qmax = 0.5 * nfft / samplingFrequency, dq = qmax / (nq - 1); autoPowerCepstrogram thee = PowerCepstrogram_create (my xmin, my xmax, nFrames, dt, t1, 0, qmax, nq, dq, 0); autoMelderProgress progress (L"Cepstrogram analysis"); for (long iframe = 1; iframe <= nFrames; iframe++) { double t = Sampled_indexToX (thee.peek(), iframe); Sound_into_Sound (sound.peek(), sframe.peek(), t - windowDuration / 2); Vector_subtractMean (sframe.peek()); Sounds_multiply (sframe.peek(), window.peek()); autoSpectrum spec = Sound_to_Spectrum (sframe.peek(), 1); // FFT yes autoPowerCepstrum cepstrum = Spectrum_to_PowerCepstrum (spec.peek()); for (long i = 1; i <= nq; i++) { thy z[i][iframe] = cepstrum -> z[1][i]; } if ((iframe % 10) == 1) { Melder_progress ((double) iframe / nFrames, L"PowerCepstrogram analysis of frame ", Melder_integer (iframe), L" out of ", Melder_integer (nFrames), L"."); } } return thee.transfer(); } catch (MelderError) { Melder_throw (me, ": no PowerCepstrogram created."); } }
Sound Sound_lengthen_overlapAdd (Sound me, double fmin, double fmax, double factor) { try { if (my ny > 1) Melder_throw (U"Overlap-add works only on mono sounds."); autoSound sound = Data_copy (me); Vector_subtractMean (sound.peek()); autoPitch pitch = Sound_to_Pitch (sound.peek(), 0.8 / fmin, fmin, fmax); autoPointProcess pulses = Sound_Pitch_to_PointProcess_cc (sound.peek(), pitch.peek()); autoPitchTier pitchTier = Pitch_to_PitchTier (pitch.peek()); autoDurationTier duration = DurationTier_create (my xmin, my xmax); RealTier_addPoint (duration.peek(), 0.5 * (my xmin + my xmax), factor); autoSound thee = Sound_Point_Pitch_Duration_to_Sound (sound.peek(), pulses.peek(), pitchTier.peek(), duration.peek(), 1.5 / fmin); return thee.transfer(); } catch (MelderError) { Melder_throw (me, U": not lengthened."); } }
Cepstrogram Sound_to_Cepstrogram (Sound me, double analysisWidth, double dt, double maximumFrequency) { try { double windowDuration = 2 * analysisWidth; /* gaussian window */ long nFrames; // Convenience: analyse the whole sound into one Cepstrogram_frame if (windowDuration > my dx * my nx) { windowDuration = my dx * my nx; } double t1, samplingFrequency = 2 * maximumFrequency; autoSound sound = Sound_resample (me, samplingFrequency, 50); Sampled_shortTermAnalysis (me, windowDuration, dt, & nFrames, & t1); autoSound sframe = Sound_createSimple (1, windowDuration, samplingFrequency); autoSound window = Sound_createGaussian (windowDuration, samplingFrequency); double qmin, qmax, dq, q1; long nq; { // laziness: find out the proper dimensions autoSpectrum spec = Sound_to_Spectrum (sframe.peek(), 1); autoCepstrum cepstrum = Spectrum_to_Cepstrum (spec.peek()); qmin = cepstrum -> xmin; qmax = cepstrum -> xmax; dq = cepstrum -> dx; q1 = cepstrum -> x1; nq = cepstrum -> nx; } autoCepstrogram thee = Cepstrogram_create (my xmin, my xmax, nFrames, dt, t1, qmin, qmax, nq, dq, q1); autoMelderProgress progress (L"Cepstrogram analysis"); for (long iframe = 1; iframe <= nFrames; iframe++) { double t = Sampled_indexToX (thee.peek(), iframe); Sound_into_Sound (sound.peek(), sframe.peek(), t - windowDuration / 2); Vector_subtractMean (sframe.peek()); Sounds_multiply (sframe.peek(), window.peek()); autoSpectrum spec = Sound_to_Spectrum (sframe.peek(), 1); autoCepstrum cepstrum = Spectrum_to_Cepstrum (spec.peek()); for (long i = 1; i <= nq; i++) { thy z[i][iframe] = cepstrum -> z[1][i]; } if ((iframe % 10) == 1) { Melder_progress ((double) iframe / nFrames, L"Cepstrogram analysis of frame ", Melder_integer (iframe), L" out of ", Melder_integer (nFrames), L"."); } } return thee.transfer(); } catch (MelderError) { Melder_throw (me, ": no Cepstrogram created."); } }
LPC LPC_and_Sound_to_LPC_robust (LPC thee, Sound me, double analysisWidth, double preEmphasisFrequency, double k, int itermax, double tol, int wantlocation) { struct huber_struct struct_huber = { 0 }; try { double t1, samplingFrequency = 1.0 / my dx, tol_svd = 0.000001; double location = 0, windowDuration = 2 * analysisWidth; /* Gaussian window */ long nFrames, frameErrorCount = 0, iter = 0; long p = thy maxnCoefficients; if (my xmin != thy xmin || my xmax != thy xmax) { Melder_throw ("Time domains differ."); } if (my dx != thy samplingPeriod) { Melder_throw ("Sampling intervals differ."); } if (floor (windowDuration / my dx) < p + 1) { Melder_throw ("Analysis window too short."); } Sampled_shortTermAnalysis (me, windowDuration, thy dx, & nFrames, & t1); if (nFrames != thy nx || t1 != thy x1) { Melder_throw ("Incorrect retrieved analysis width"); } autoSound sound = Data_copy (me); autoSound sframe = Sound_createSimple (1, windowDuration, samplingFrequency); autoSound window = Sound_createGaussian (windowDuration, samplingFrequency); autoLPC him = Data_copy (thee); huber_struct_init (&struct_huber, windowDuration, p, samplingFrequency, location, wantlocation); struct_huber.k = k; struct_huber.tol = tol; struct_huber.tol_svd = tol_svd; struct_huber.itermax = itermax; autoMelderProgress progess (L"LPC analysis"); Sound_preEmphasis (sound.peek(), preEmphasisFrequency); for (long i = 1; i <= nFrames; i++) { LPC_Frame lpc = (LPC_Frame) & thy d_frames[i]; LPC_Frame lpcto = (LPC_Frame) & his d_frames[i]; double t = Sampled_indexToX (thee, i); Sound_into_Sound (sound.peek(), sframe.peek(), t - windowDuration / 2); Vector_subtractMean (sframe.peek()); Sounds_multiply (sframe.peek(), window.peek()); try { LPC_Frames_and_Sound_huber (lpc, sframe.peek(), lpcto, & struct_huber); } catch (MelderError) { frameErrorCount++; } iter += struct_huber.iter; if ( (i % 10) == 1) { Melder_progress ( (double) i / nFrames, L"LPC analysis of frame ", Melder_integer (i), L" out of ", Melder_integer (nFrames), L"."); } } if (frameErrorCount) Melder_warning (L"Results of ", Melder_integer (frameErrorCount), L" frame(s) out of ", Melder_integer (nFrames), L" could not be optimised."); MelderInfo_writeLine4 (L"Number of iterations: ", Melder_integer (iter), L"\n Average per frame: ", Melder_double (((double) iter) / nFrames)); huber_struct_destroy (&struct_huber); return him.transfer(); } catch (MelderError) { huber_struct_destroy (&struct_huber); Melder_throw (me, ": no robust LPC created."); } }