[1.39.28]

Added experimental Klatt synthesizer.
Klatt phonemes use a new format of phoneme data.
espeakedit, new dialog fields and layout for Klatt phonemes.
Added Klatt versions of English phonemes.

Fixed some compilation warnings on Windows.
Fixed bug in Lettergroups > L10

Increased amplitude of harmonics below F1 peak, to increase bass.
Allow 5 (up from 4) pairs of values in "tone" command in voice files.
Default to slightly wider pk_shape1.

If words are separated by hyphens, don't add word-gap, and don't pass new-word flag onto phoneme processing.


git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@203 d46cf337-b52f-0410-862d-fd96e6ae7743

src/compiledata.cpp

@@ -597,17 +597,21 @@ int Compile::LoadSpect(const char *path, int control)
 	int peak;
 	int displ;
 	int frame;
+	int n_frames;
 	int ix;
-	int x;
+	int x, x2;
 	int rms;
 	float total;
 	float pkheight;
 	int marker1_set=0;
 	int frame_vowelbreak=0;
+	int klatt_flag=0;
 	SpectFrame *fr;
+	frame_t *fr_out;
 	wxString path_sep = _T("/");
 
 	SPECT_SEQ seq_out;
+	SPECT_SEQK seqk_out;
 
 	// create SpectSeq and import data
 	spectseq = new SpectSeq;
@@ -627,11 +631,14 @@ int Compile::LoadSpect(const char *path, int control)
 	}
 	spectseq->Load(stream);
 
+	if(spectseq->synthesizer_type == 1)
+		klatt_flag = FRFLAG_KLATT;
+
 	displ = ftell(f_phdata);
 
 	seq_out.n_frames=0;
 	seq_out.flags=0;
-	seq_out.length=0;
+	seq_out.length_total=0;
 
 	total = 0;	
 	for(frame=0; frame < spectseq->numframes; frame++)
@@ -669,7 +676,7 @@ for(ix=0; ix<8; ix++)
 				total += spectseq->frames[frame-1]->length;
 		}	
 	}
-	seq_out.length = int(total);
+	seq_out.length_total = int(total);
 
 	if((control & 1) && (marker1_set == 0))
 	{
@@ -678,68 +685,103 @@ for(ix=0; ix<8; ix++)
 		spectseq->frames[frame_vowelbreak]->markers |= FRFLAG_VOWEL_CENTRE;
 	}
 
-	ix = 0;
+	n_frames = 0;
 	for(frame=0; frame < spectseq->numframes; frame++)
 	{
 		fr = spectseq->frames[frame];
 		
 		if(fr->keyframe)
 		{
+			if(klatt_flag)
+				fr_out = &seqk_out.frame[n_frames];
+			else
+				fr_out = (frame_t *)&seq_out.frame[n_frames];
+
 			x = int(fr->length + 0.5);  // round to nearest mS
 			if(x > 255) x = 255;
-			seq_out.frame[ix].length = x;
+			fr_out->length = x;
+
+			fr_out->frflags = fr->markers | klatt_flag;
 
-			seq_out.frame[ix].frflags = fr->markers;
 			rms = int(fr->GetRms(spectseq->amplitude));
 			if(rms > 255) rms = 255;
-			seq_out.frame[ix].rms = rms;
+			fr_out->rms = rms;
+
+			if(n_frames == (seq_out.n_frames-1))
+				fr_out->length = 0;    // give last frame zero length
 
-			if(ix == (seq_out.n_frames-1))
-				seq_out.frame[ix].length = 0;    // give last frame zero length
-				
 			// write: peak data
 			count_frames++;
 			for(peak=0; peak<N_PEAKS; peak++)
 			{
-				seq_out.frame[ix].ffreq[peak] = fr->peaks[peak].pkfreq;
-				pkheight = spectseq->amplitude * fr->amp_adjust * fr->peaks[peak].pkheight;
-				pkheight = pkheight/640000;
-				if(pkheight > 255) pkheight = 255;
-				seq_out.frame[ix].fheight[peak] = int(pkheight);
+				fr_out->ffreq[peak] = fr->peaks[peak].pkfreq;
+
+				if(klatt_flag)
+				{
+					pkheight = fr->peaks[peak].pkheight / 128;
+				}
+				else
+				{
+					pkheight = spectseq->amplitude * fr->amp_adjust * fr->peaks[peak].pkheight;
+					pkheight = pkheight/640000;
+					if(pkheight > 255) pkheight = 255;
+				}
+				fr_out->fheight[peak] = int(pkheight);
 
 				if(peak < 6)
 				{
-					x =  fr->peaks[peak].pkwidth/4;
+					x =  fr->peaks[peak].pkwidth/2;
+					x2 =  fr->peaks[peak].pkright;
+					if(klatt_flag == 0)
+					{
+						x /= 2;
+						x2 /= 4;
+					}
 					if(x > 255) x = 255;
-					seq_out.frame[ix].fwidth[peak] = x;
+					fr_out->fwidth[peak] = x;
 
-					x =  fr->peaks[peak].pkright/4;
-					if(x > 255) x = 255;
-					seq_out.frame[ix].fright[peak] = x;
+					if(x2 > 255) x2 = 255;
+					fr_out->fright[peak] = x2;
 				}
 			}
 
-#ifdef LOG_DETAIL
-fprintf(f_errors,"Frame %d  %3dmS  rms=%3d  flags=%2d  pk=%4d %4d %4d",ix,seq_out.frame[ix].length,
-	seq_out.frame[ix].rms,seq_out.frame[ix].flags,
-	seq_out.frame[ix].peaks[1].pkfreq,seq_out.frame[ix].peaks[2].pkfreq,seq_out.frame[ix].peaks[3].pkfreq);
+			if(klatt_flag)
+			{
+				if(fr->klatt_param[KLATT_Kopen] == 0)
+				{
+					Error("Klatt_Kopen has value 0");
+				}
 
-	if(fr->markers != 0)
-	{
-		fprintf(f_errors," [%x]",fr->markers);
-	}
-	fputc('\n',f_errors);
-#endif
-	
-			ix++;
+				// additional fields
+				for(ix=0; ix<N_KLATTP; ix++)
+				{
+					fr_out->klattp[ix] = fr->klatt_param[ix];
+				}
+				fr_out->fwidth6 = fr->peaks[6].pkwidth/2;
+				fr_out->fright6 = fr->peaks[6].pkright;
+			}
+
+			n_frames++;
 		}
 	}
 	
-	ix = (char *)(&seq_out.frame[seq_out.n_frames]) - (char *)(&seq_out);
-	ix = (ix+3) & 0xfffc;   // round up to multiple of 4 bytes
+	if(klatt_flag)
+	{
+		seqk_out.n_frames = seq_out.n_frames;
+		seqk_out.flags = seq_out.flags;
+		seqk_out.length_total = seq_out.length_total;
+
+		ix = (char *)(&seqk_out.frame[seqk_out.n_frames]) - (char *)(&seqk_out);
+		ix = (ix+3) & 0xfffc;   // round up to multiple of 4 bytes
+		fwrite(&seqk_out,ix,1,f_phdata);
+	}
+	else
+	{
+		ix = (char *)(&seq_out.frame[seq_out.n_frames]) - (char *)(&seq_out);
+		ix = (ix+3) & 0xfffc;   // round up to multiple of 4 bytes
+		fwrite(&seq_out,ix,1,f_phdata);
+	}
 
-	fwrite(&seq_out,ix,1,f_phdata);
-	
 	delete spectseq;
 	return(displ);
 }  //  end of Compile::LoadSpect

src/compiledict.cpp

@@ -1268,7 +1268,7 @@ static int compile_lettergroup(char *input, FILE *f_out)
 	group = atoi(&p[0]);
 	if(group >= N_LETTER_GROUPS)
 	{
-		fprintf(f_log,"%5d: lettergroup out of range (01-%.2d)\n",linenum,N_LETTER_GROUPS);
+		fprintf(f_log,"%5d: lettergroup out of range (01-%.2d)\n",linenum,N_LETTER_GROUPS-1);
 		error_count++;
 		return(1);
 	}

src/espeakedit.cpp

@@ -232,7 +232,7 @@ wxSashLayoutWindow *win;
   win = new wxSashLayoutWindow(this, ID_WINDOW_LEFT2,
                                wxDefaultPosition, wxSize(200, 30),
                                wxNO_BORDER | wxSW_3D | wxCLIP_CHILDREN);
-  win->SetDefaultSize(wxSize(208, 1000));
+  win->SetDefaultSize(wxSize(212, 1000));
   win->SetOrientation(wxLAYOUT_VERTICAL);
   win->SetAlignment(wxLAYOUT_LEFT);
 //  win->SetBackgroundColour(wxColour(0, 255, 255));

src/formantdlg.cpp

@@ -43,10 +43,10 @@ ByteGraph *pitchgraph=NULL;
 BEGIN_EVENT_TABLE(FormantDlg, wxPanel)
 	EVT_BUTTON(T_ZOOMOUT,FormantDlg::OnCommand)
 	EVT_BUTTON(T_ZOOMIN,FormantDlg::OnCommand)
-	EVT_SPINCTRL(T_AMPLITUDE,FormantDlg::OnSpin)
-	EVT_SPINCTRL(T_TIMESEQ,FormantDlg::OnSpin)
-	EVT_SPINCTRL(T_AMPFRAME,FormantDlg::OnSpin)
-	EVT_SPINCTRL(T_TIMEFRAME,FormantDlg::OnSpin)
+	EVT_SPINCTRL(-1,FormantDlg::OnSpin)
+//	EVT_SPINCTRL(T_TIMESEQ,FormantDlg::OnSpin)
+//	EVT_SPINCTRL(T_AMPFRAME,FormantDlg::OnSpin)
+//	EVT_SPINCTRL(T_TIMEFRAME,FormantDlg::OnSpin)
 END_EVENT_TABLE()
 
 
@@ -79,21 +79,19 @@ void FormantDlg::OnSpin(wxSpinEvent& event)
 
 	switch(id = event.GetId())
 	{
-	case T_AMPLITUDE:
-	case T_TIMESEQ:
-	case T_TIMEFRAME:
-		currentcanvas->RefreshDialogValues(0);
-		break;
-
 	case T_AMPFRAME:
 		currentcanvas->RefreshDialogValues(1);
 		break;
+
+	default:
+		currentcanvas->RefreshDialogValues(0);
+		break;
 	}
 	currentcanvas->SetFocus();
 }
 
 FormantDlg::FormantDlg(wxWindow *parent) : wxPanel(parent,-1,wxDefaultPosition,wxSize(400,1000))
-{//==========================================================================
+{//=============================================================================================
 	int  ix;
 	int  y;
 	int  xplace;
@@ -104,6 +102,7 @@ FormantDlg::FormantDlg(wxWindow *parent) : wxPanel(parent,-1,wxDefaultPosition,w
 	xplace = 28;
 	id = 201;
 	new wxStaticText(this,-1,_T("Formants"),wxPoint(4,5));
+	t_lab[4] = new wxStaticText(this,-1,_T("  BW      BWp      Ap"),wxPoint(78,5),wxSize(300,20));
 
 	for(ix=0; ix<N_PEAKS; ix++)
 	{
@@ -115,48 +114,124 @@ FormantDlg::FormantDlg(wxWindow *parent) : wxPanel(parent,-1,wxDefaultPosition,w
 			wxPoint(xplace,y+24*ix),wxSize(48,20),
 			wxTE_CENTRE+wxTE_READONLY);
 		t_pkheight[ix] = new wxTextCtrl(this,id++,_T(""),
-			wxPoint(xplace+52,y+24*ix),wxSize(36,20),
+			wxPoint(xplace+50,y+24*ix),wxSize(36,20),
 			wxTE_CENTRE+wxTE_READONLY);
 		t_pkwidth[ix] = new wxTextCtrl(this,id++,_T(""),
-			wxPoint(xplace+92,y+24*ix),wxSize(48,20),
+			wxPoint(xplace+88,y+24*ix),wxSize(48,20),
 			wxTE_CENTRE+wxTE_READONLY);
-#ifdef WIDTH2
 		t_pkright[ix] = new wxTextCtrl(this,id++,_T(""),
-			wxPoint(xplace+132,y+24*ix),wxSize(36,20),
+			wxPoint(xplace+138,y+24*ix),wxSize(36,20),
 			wxTE_CENTRE+wxTE_READONLY);
-#endif
 		if(ix>5) t_pkwidth[ix]->Hide();
 	}
 
-	t_ampframe = new wxSpinCtrl(this,T_AMPFRAME,_T(""),
-		wxPoint(6,244),wxSize(52,24),wxTE_CENTRE,0,500);
-	t_lab[3] = new wxStaticText(this,-1,_T("% amp - Frame"),wxPoint(61,248));
-	t_timeframe = new wxSpinCtrl(this,T_TIMEFRAME,_T(""),
-		wxPoint(6,272),wxSize(52,24),wxTE_CENTRE,0,500);
-	t_orig_frame = new wxStaticText(this,-1,_T("mS"),wxPoint(61,276));
+	y=240;
+	t_timeframe = new wxSpinCtrl(this,T_TIMEFRAME,_T(""), wxPoint(6,y+0), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_orig_frame = new wxStaticText(this,-1,_T("mS"),wxPoint(57,y+8));
+	t_ampframe = new wxSpinCtrl(this,T_AMPFRAME,_T(""), wxPoint(100,y+0), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_lab[3] = new wxStaticText(this,-1,_T("% amp - Frame"),wxPoint(151,y+8));
 
-	t_zoomout = new wxButton(this,T_ZOOMOUT,_T("Zoom-"),wxPoint(16,310));
-	t_zoomin = new wxButton(this,T_ZOOMIN,_T("Zoom+"),wxPoint(106,310));
+	s_klatt[KLATT_AV] = new wxSpinCtrl(this,T_AV,_T(""), wxPoint(6,y+32), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_AV] = new wxStaticText(this,-1,_T("AV"),wxPoint(61,y+36));
+	s_klatt[KLATT_AVp] = new wxSpinCtrl(this,T_AVP,_T(""), wxPoint(100,y+32), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_AVp] = new wxStaticText(this,-1,_T("AVp"),wxPoint(155,y+36));
+
+	s_klatt[KLATT_Fric] = new wxSpinCtrl(this,T_FRIC,_T(""), wxPoint(6,y+60), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_Fric] = new wxStaticText(this,-1,_T("Fric"),wxPoint(61,y+64));
+	s_klatt[KLATT_FricBP] = new wxSpinCtrl(this,T_FRICBP,_T(""), wxPoint(100,y+60), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_FricBP] = new wxStaticText(this,-1,_T("FricBP"),wxPoint(155,y+64));
+
+	s_klatt[KLATT_Aspr] = new wxSpinCtrl(this,T_ASPR,_T(""), wxPoint(6,y+88), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_Aspr] = new wxStaticText(this,-1,_T("Aspr"),wxPoint(61,y+92));
+	s_klatt[KLATT_Turb] = new wxSpinCtrl(this,T_TURB,_T(""), wxPoint(100,y+88), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_Turb] = new wxStaticText(this,-1,_T("Turb"),wxPoint(155,y+92));
+
+	s_klatt[KLATT_Skew] = new wxSpinCtrl(this,T_SKEW,_T(""), wxPoint(6,y+116), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_Skew] = new wxStaticText(this,-1,_T("Skew"),wxPoint(61,y+120));
+	s_klatt[KLATT_Tilt] = new wxSpinCtrl(this,T_TILT,_T(""), wxPoint(100,y+116), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_Tilt] = new wxStaticText(this,-1,_T("Tilt"),wxPoint(155,y+120));
+
+	s_klatt[KLATT_Kopen] = new wxSpinCtrl(this,T_KOPEN,_T(""), wxPoint(6,y+144), wxSize(52,24), wxTE_CENTRE,0,500);
+	t_klatt[KLATT_Kopen] = new wxStaticText(this,-1,_T("kopen"),wxPoint(61,y+148));
 
 
-//	usepitchenv = new wxCheckBox(this,T_USEPITCHENV,_T("???"),wxPoint(20,320));
+	t_zoomout = new wxButton(this,T_ZOOMOUT,_T("Zoom-"),wxPoint(16,420));
+	t_zoomin = new wxButton(this,T_ZOOMIN,_T("Zoom+"),wxPoint(106,420));
 
+	y=468;
 	t_amplitude = new wxSpinCtrl(this,T_AMPLITUDE,_T(""),
-		wxPoint(6,370),wxSize(52,24),wxTE_CENTRE,0,500);
-	t_lab[2] = new wxStaticText(this,-1,_T("% amp - Sequence"),wxPoint(61,374));
+		wxPoint(6,y),wxSize(52,24),wxTE_CENTRE,0,y+130);
+	t_lab[2] = new wxStaticText(this,-1,_T("% amp - Sequence"),wxPoint(61,y+4));
 //	t_timeseq = new wxSpinCtrl(this,T_TIMESEQ,_T(""),
 //		wxPoint(6,400),wxSize(52,24),wxTE_CENTRE,0,500);
-	t_orig_seq = new wxStaticText(this,-1,_T("mS"),wxPoint(61,404));
+	t_orig_seq = new wxStaticText(this,-1,_T("mS"),wxPoint(61,y+30));
 
-	t_pitch = new wxStaticText(this,-1,_T(""),wxPoint(4,440),wxSize(192,24));
+	t_pitch = new wxStaticText(this,-1,_T(""),wxPoint(4,520),wxSize(192,24));
 
-	pitchgraph = new ByteGraph(this,wxPoint(0,460),wxSize(200,140));
+	pitchgraph = new ByteGraph(this,wxPoint(0,538),wxSize(200,140));
 	pitchgraph->SetData(128,env_fall);
 	pitchgraph->ShowSpectrum(1);
 	pitchgraph->Show();
 }
 
 
+void FormantDlg::HideFields(int synth_type)
+{//=======================================
+	int ix;
+	static int prev_synth_type = -1;
+
+	if(synth_type == prev_synth_type)
+		return;  // no change
+
+	if(synth_type == 0)
+	{
+		for(ix=0; ix<9; ix++)
+		{
+			s_klatt[ix]->Hide();
+			t_klatt[ix]->Hide();
+		}
+		for(ix=6; ix<9; ix++)
+		{
+			t_pkfreq[ix]->Show();
+			t_pkheight[ix]->Show();
+			t_pkwidth[ix]->Hide();
+			t_pkright[ix]->Hide();
+		}
+		t_ampframe->Show();
+		t_amplitude->Show();
+		t_lab[2]->Show();
+		t_lab[3]->Show();
+		t_labpk[7]->Show();
+		t_labpk[8]->Show();
+		t_lab[4]->Hide();
+	}
+	else
+	{
+		for(ix=0; ix<9; ix++)
+		{
+			s_klatt[ix]->Show();
+			t_klatt[ix]->Show();
+		}
+		for(ix=7; ix<9; ix++)
+		{
+			t_pkfreq[ix]->Hide();
+			t_pkheight[ix]->Hide();
+			t_pkwidth[ix]->Hide();
+			t_pkright[ix]->Hide();
+		}
+		t_pkwidth[6]->Show(1);
+		t_pkright[6]->Show(1);
+		t_ampframe->Hide();
+		t_amplitude->Hide();
+		t_lab[2]->Hide();
+		t_lab[3]->Hide();
+		t_labpk[7]->Hide();
+		t_labpk[8]->Hide();
+		t_lab[4]->Show();
+	}
+	prev_synth_type = synth_type;
+}
+
 
 void FormantDlg::ShowFrame(SpectSeq *spectseq, int frame, int pk, int field)
 {//==============================================================
@@ -169,6 +244,8 @@ void FormantDlg::ShowFrame(SpectSeq *spectseq, int frame, int pk, int field)
 		return;
 	sf = spectseq->frames[frame];
 
+	HideFields(spectseq->synthesizer_type);
+
 	if(field == 0xff)
 	{
 		// indicate the selected peak
@@ -203,10 +280,8 @@ void FormantDlg::ShowFrame(SpectSeq *spectseq, int frame, int pk, int field)
 		{
 			value.Printf(_T("%3d"),sf->peaks[ix].pkwidth);
 			t_pkwidth[ix]->SetValue(value);
-#ifdef WIDTH2
 			value.Printf(_T("%3d"),sf->peaks[ix].pkright);
 			t_pkright[ix]->SetValue(value);
-#endif
 		}
 	}
 
@@ -216,6 +291,16 @@ void FormantDlg::ShowFrame(SpectSeq *spectseq, int frame, int pk, int field)
 	t_orig_frame->SetLabel(value);
 	value.Printf(_T("%3d"),sf->amp_adjust);
 	t_ampframe->SetValue(value);
+
+	if(spectseq->synthesizer_type == 1)
+	{
+		for(ix=0; ix<9; ix++)
+		{
+			SetSpinCtrl(s_klatt[ix], sf->klatt_param[ix]);
+		}
+	}
+
+
 }  //  end of FormantDlg::ShowFrame
 
 

src/klatt.h

@@ -0,0 +1,138 @@
+
+
+#define CASCADE_PARALLEL 1         /* Type of synthesis model */
+#define ALL_PARALLEL     2 
+
+#define IMPULSIVE        1         /* Type of voicing source */
+#define NATURAL          2
+#define SAMPLED          3
+
+#define PI               3.1415927
+
+
+/* typedef's that need to be exported */
+
+typedef long flag;
+
+/* Resonator Structure */
+
+typedef struct
+{
+	double a;
+	double b;
+	double c;
+	double p1;
+	double p2;
+	double a_inc;
+	double b_inc;
+	double c_inc;
+} resonator_t, *resonator_ptr;
+
+/* Structure for Klatt Globals */
+
+typedef struct
+{
+  flag synthesis_model; /* cascade-parallel or all-parallel */
+  flag outsl;       /* Output waveform selector                      */
+  long samrate;     /* Number of output samples per second           */
+  long FLPhz ;      /* Frequeny of glottal downsample low-pass filter */
+  long BLPhz ;      /* Bandwidth of glottal downsample low-pass filter */
+  flag glsource;    /* Type of glottal source */
+  int f0_flutter;   /* Percentage of f0 flutter 0-100 */
+  long nspfr;       /* number of samples per frame */
+  long nper;        /* Counter for number of samples in a pitch period */
+  long ns;
+  long T0;          /* Fundamental period in output samples times 4 */
+  long nopen;       /* Number of samples in open phase of period    */
+  long nmod;        /* Position in period to begin noise amp. modul */
+  long nrand;       /* Varible used by random number generator      */
+  double pulse_shape_a;  /* Makes waveshape of glottal pulse when open   */
+  double pulse_shape_b;  /* Makes waveshape of glottal pulse when open   */
+  double minus_pi_t;
+  double two_pi_t;
+  double onemd;
+  double decay;
+  double amp_bypas; /* AB converted to linear gain              */
+  double amp_voice; /* AVdb converted to linear gain            */
+  double par_amp_voice; /* AVpdb converted to linear gain       */
+  double amp_aspir; /* AP converted to linear gain              */
+  double amp_frica; /* AF converted to linear gain              */
+  double amp_breth; /* ATURB converted to linear gain           */
+  double amp_gain0; /* G0 converted to linear gain              */
+  int num_samples; /* number of glottal samples */
+  double sample_factor; /* multiplication factor for glottal samples */
+  short *natural_samples; /* pointer to an array of glottal samples */
+  long original_f0; /* original value of f0 not modified by flutter */
+
+	int fadeout;       // set to 64 to cause fadeout over 64 samples
+	int scale_wav;     // depends on the voicing source
+
+#define N_RSN 20
+#define Rnpc 0
+#define R1c  1
+#define R2c  2
+#define R3c  3
+#define R4c  4
+#define R5c  5
+#define R6c  6
+#define R7c  7
+#define R8c  8
+#define Rnz  9
+
+#define Rparallel 10
+#define Rnpp 10
+#define R1p  11
+#define R2p  12
+#define R3p  13
+#define R4p  14
+#define R5p  15
+#define R6p  16
+
+#define RGL  17
+#define RLP  18
+#define Rout 19
+
+  resonator_t rsn[N_RSN];	 // internal storage for resonators
+  resonator_t rsn_next[N_RSN];
+
+} klatt_global_t, *klatt_global_ptr;
+  
+/* Structure for Klatt Parameters */
+
+#define F_NP   0  // nasal zero formant
+#define F1     1
+#define F2     2
+#define F3     3
+#define F4     4
+#define F5     5
+#define F6     6
+#define F_NZ   9  // nasal pole formant
+
+
+typedef struct
+{
+	long F0hz10; /* Voicing fund freq in Hz                          */
+	long AVdb;   /* Amp of voicing in dB,            0 to   70       */
+	int Fhz[10];  // formant Hz, F_NZ to F6 to F_NP
+	int Bhz[10];
+	int Ap[10];   /* Amp of parallel formants in dB,    0 to   80       */
+	int Bphz[10]; /* Parallel formants bw in Hz,       40 to 1000      */
+
+	long ASP;    /* Amp of aspiration in dB,         0 to   70       */
+	long Kopen;  /* # of samples in open period,     10 to   65      */
+	long Aturb;  /* Breathiness in voicing,          0 to   80       */
+	long TLTdb;  /* Voicing spectral tilt in dB,     0 to   24       */
+	long AF;     /* Amp of frication in dB,          0 to   80       */
+	long Kskew;  /* Skewness of alternate periods,   0 to   40 in sample#/2  */
+
+	long AB;     /* Amp of bypass fric. in dB,       0 to   80       */
+	long AVpdb;  /* Amp of voicing,  par in dB,      0 to   70       */
+	long Gain0;  /* Overall gain, 60 dB is unity,    0 to   60       */
+
+	long AVdb_tmp;      //copy of AVdb, which is changed within parwave()
+	int Fhz_next[10];    // Fhz for the next chunk, so we can do interpolation of resonator (a,b,c) parameters
+	int Bhz_next[10];
+ } klatt_frame_t, *klatt_frame_ptr;
+
+
+

src/readclause.cpp

@@ -1186,7 +1186,7 @@ static int attr_prosody_value(int param_type, const wchar_t *pw, int *value_out)
 		pw++;	
 		sign = -1;
 	}
-	value = wcstod(pw,&tail);
+	value = (float)wcstod(pw,&tail);
 	if(tail == pw)
 	{
 		// failed to find a number, return 100%

src/speak_lib.cpp

@@ -300,10 +300,10 @@ static void init_path(const char *path)
 	}
 
 	buf[0] = 0;
-	RegOpenKeyEx(HKEY_LOCAL_MACHINE, "Software\\Microsoft\\Speech\\Voices\\Tokens\\eSpeak", 0, KEY_READ, &RegKey);
+	RegOpenKeyExA(HKEY_LOCAL_MACHINE, "Software\\Microsoft\\Speech\\Voices\\Tokens\\eSpeak", 0, KEY_READ, &RegKey);
 	size = sizeof(buf);
 	var_type = REG_SZ;
-	RegQueryValueEx(RegKey, "path", 0, &var_type, buf, &size);
+	RegQueryValueExA(RegKey, "path", 0, &var_type, buf, &size);
 
 	sprintf(path_home,"%s\\espeak-data",buf);
 

src/spect.cpp

@@ -137,7 +137,8 @@ int SpectFrame::Import(wxInputStream& stream1)
 	double x;
 	unsigned short *spect_data;
 
-   wxTextInputStream stream(stream1);
+	synthesizer_type = 0;
+	wxTextInputStream stream(stream1);
 
 	stream >> time;
 	stream >> pitch;
@@ -273,12 +274,13 @@ int SpectFrame::ImportSPC2(wxInputStream& stream, float &time_acc)
 
 
 
-int SpectFrame::Load(wxInputStream& stream)
-{//========================================
+int SpectFrame::Load(wxInputStream& stream, int synth_type)
+{//========================================================
 	int  ix;
 	int  x;
 	unsigned short *spect_data;
 
+	synthesizer_type = synth_type;
    wxDataInputStream s(stream);
 
 	time = s.ReadDouble();
@@ -300,6 +302,14 @@ int SpectFrame::Load(wxInputStream& stream)
 		peaks[ix].pkright = s.Read16();
 	}
 
+	if(synthesizer_type == 1)
+	{
+		for(ix=0; ix<N_KLATTP; ix++)
+		{
+			klatt_param[ix] = s.Read16();
+		}
+	}
+
 	spect_data = new USHORT[nx];
 
 	if(spect_data == NULL)
@@ -324,7 +334,7 @@ int SpectFrame::Save(wxOutputStream& stream)
 {//=========================================
 	int ix;
 
-   wxDataOutputStream s(stream);
+  	wxDataOutputStream s(stream);
 
 	s.WriteDouble(time);
 	s.WriteDouble(pitch);
@@ -345,6 +355,14 @@ int SpectFrame::Save(wxOutputStream& stream)
 		s.Write16(peaks[ix].pkright);
 	}
 
+	if(synthesizer_type == 1)
+	{
+		for(ix=0; ix<N_KLATTP; ix++)
+		{
+			s.Write16(klatt_param[ix]);
+		}
+	}
+
 	for(ix=0; ix<nx; ix++)
 	{
 		s.Write16(spect[ix]);
@@ -367,6 +385,7 @@ void SpectFrame::ZeroPeaks()
 void SpectFrame::CopyPeaks(SpectFrame *sf)
 {//=======================================
 	memcpy(peaks,sf->peaks,sizeof(peaks));
+	memcpy(klatt_param, sf->klatt_param, sizeof(klatt_param));
 	keyframe = sf->keyframe;
 }
 
@@ -448,6 +467,9 @@ void SpectFrame::DrawPeaks(wxDC *dc, int offy, int frame_width, int seq_amplitud
 	int  y1, y2;
 	double yy;
 	int max_ix;
+	int height;
+	int pkright;
+	int pkwidth;
 	int buf[DRAWPEAKWIDTH*2];
 
 	max_ix = int(9000 * scale_x);
@@ -459,9 +481,24 @@ void SpectFrame::DrawPeaks(wxDC *dc, int offy, int frame_width, int seq_amplitud
 
 		if((pk->pkfreq == 0) || (pk->pkheight==0)) continue;
 
+		if(synthesizer_type==1)
+		{
+			if(peak==0) continue;
+
+			height = pk->pkheight + (64 * 64);
+			height = (500000/height) * 100;
+			pkright = pkwidth = pk->pkheight / 16;
+		}
+		else
+		{
+			height = pk->pkheight;
+			pkright = pk->pkright;
+			pkwidth = pk->pkwidth;
+		}
+
 		x1 = (int)(pk->pkfreq*scale_x);
-		x2 = (int)((pk->pkfreq + pk->pkright)*scale_x);
-		x3 = (int)((pk->pkfreq - pk->pkwidth)*scale_x);
+		x2 = (int)((pk->pkfreq + pkright)*scale_x);
+		x3 = (int)((pk->pkfreq - pkwidth)*scale_x);
 
 		if(x3 >= DRAWPEAKWIDTH)
 			continue;   // whole peak is off the scale
@@ -469,7 +506,7 @@ void SpectFrame::DrawPeaks(wxDC *dc, int offy, int frame_width, int seq_amplitud
 		if((width = x2-x1) <= 0) continue;
 		for(ix=0; ix<width; ix++)
 		{
-			buf[x1+ix] += pk->pkheight * pk_shape1[(ix*PEAKSHAPEW)/width];
+			buf[x1+ix] += height * pk_shape1[(ix*PEAKSHAPEW)/width];
 		}
 
 		if((width = x1-x3) <= 0) continue;
@@ -477,7 +514,7 @@ void SpectFrame::DrawPeaks(wxDC *dc, int offy, int frame_width, int seq_amplitud
 		{
 			if(x3+ix >= 0)
 			{
-				buf[x3+ix] += pk->pkheight * pk_shape1[((width-ix)*PEAKSHAPEW)/width];
+				buf[x3+ix] += height * pk_shape1[((width-ix)*PEAKSHAPEW)/width];
 			}
 		}
 	}
@@ -633,3 +670,71 @@ void SpectFrame::Draw(wxDC& dc, int offy, int frame_width, double scalex, double
 
 }  // end of SpectFrame::Draw
 
+
+void SpectFrame::KlattDefaults()
+{//============================
+	// set default values for Klatt parameters
+	int pk;
+	int bw;
+	int bw3;
+
+	synthesizer_type = 1;
+
+	klatt_param[KLATT_AV] = 59;
+	klatt_param[KLATT_AVp] = 0;
+	klatt_param[KLATT_Fric] = 0;
+	klatt_param[KLATT_FricBP] = 0;
+	klatt_param[KLATT_Aspr] = 0;
+	klatt_param[KLATT_Turb] = 0;
+	klatt_param[KLATT_Skew] = 0;
+	klatt_param[KLATT_Tilt] = 0;
+	klatt_param[KLATT_Kopen] = 40;
+	klatt_param[KLATT_spare1] = 0;
+
+	bw = 60;
+	if(peaks[1].pkfreq < 400)
+		bw = 55;
+	if(peaks[1].pkfreq > 600)
+		bw = 70;
+	if(peaks[1].pkfreq > 650)
+		bw = 80;
+	if(peaks[1].pkfreq > 750)
+		bw = 90;
+	peaks[1].pkwidth = bw;
+
+	bw = 90;
+	bw3 = 150;
+	if(peaks[2].pkfreq < 1000)
+	{
+		bw = 80;
+		bw3 = 120;
+	}
+	if(peaks[2].pkfreq > 1600)
+	{
+		bw = 100;
+		bw3 = 200;
+	}
+	if(peaks[2].pkfreq > 2000)
+	{
+		bw = 110;
+		bw3 = 250;
+	}
+	peaks[2].pkwidth = bw;
+	peaks[3].pkwidth = bw3;
+	peaks[4].pkwidth = 200;
+	peaks[5].pkwidth = 200;
+	peaks[6].pkwidth = 500;
+	peaks[0].pkfreq = 280;   // FNP
+	peaks[0].pkwidth = 280; // FNZ
+
+	peaks[7].pkfreq = 7800;
+	peaks[7].pkwidth = 0;
+	peaks[8].pkfreq = 9000;
+	peaks[8].pkwidth = 0;
+
+	for(pk=0; pk<=8; pk++)
+	{
+		peaks[pk].pkheight = peaks[pk].pkwidth << 6;
+		peaks[pk].pkright = 0;
+	}
+}

src/spect.h

@@ -39,8 +39,20 @@
 #define T_TIMEFRAME 310
 #define T_TIMESEQ   311
 
+#define T_AV      312
+#define T_AVP     313
+#define T_FRIC    314
+#define T_FRICBP  315
+#define T_ASPR    316
+#define T_TURB    317
+#define T_SKEW    318
+#define T_TILT    319
+#define T_KOPEN   320
+
+
 #define FILEID1_SPECTSEQ 0x43455053
-#define FILEID2_SPECTSEQ 0x51455354
+#define FILEID2_SPECTSEQ 0x51455354  // for eSpeak sequence
+#define FILEID2_SPECTSEK 0x4b455354  // for Klatt sequence
 
 #define FILEID1_SPC2     0x32435053  // an old format for spectrum files
 
@@ -110,7 +122,7 @@ public:
 	~SpectFrame();
    int Import(wxInputStream &stream);
    int ImportSPC2(wxInputStream &stream, float &time_acc);
-	int Load(wxInputStream &stream);
+	int Load(wxInputStream &stream, int synthesizer_type);
 	int Save(wxOutputStream &stream);
    void Draw(wxDC &dc, int offy, int frame_width, double scalex, double scaley);
 
@@ -122,6 +134,7 @@ public:
 	void MakeWave(int peaks, PitchEnvelope &pitch, int amplitude, int duration);
 	void MakeHtab(int numh, int *htab, int pitch);
 	double GetRms(int amp);
+	void KlattDefaults();
 
 	int selected;
 	int keyframe;
@@ -136,8 +149,11 @@ public:
    int  nx;
    int  markers;
 	int  max_y;
+	int  synthesizer_type;  // 0=eSpeak, 1=Klatt
    USHORT *spect;    // sqrt of harmonic amplitudes,  1-nx at 'pitch'
 
+	short  klatt_param[N_KLATTP];
+
    formant_t formants[N_PEAKS];   // this is just the estimate given by Praat
    peak_t peaks[N_PEAKS];
 
@@ -171,7 +187,8 @@ public:
 	float GetKeyedLength();
 	void SetFrameLengths();
 	void ConstructVowel(void);
-      
+	void SetKlattDefaults(void);
+
 	int  numframes;
 	int  amplitude;
 	int  spare;
@@ -186,6 +203,8 @@ public:
 	int  bass_reduction;
 	int max_x;
 	int max_y;
+	int synthesizer_type;
+
 
 private:
 	void Load2(wxInputStream& stream, int import, int n);
@@ -262,8 +281,9 @@ class FormantDlg : public wxPanel
 	public:
 		FormantDlg(wxWindow *parent);
 		void ShowFrame(SpectSeq *spectseq, int frame, int pk, int field);
-      void OnCommand(wxCommandEvent& event);
-      void OnSpin(wxSpinEvent& event);
+		void OnCommand(wxCommandEvent& event);
+		void OnSpin(wxSpinEvent& event);
+		void HideFields(int synth_type);
 
 		wxCheckBox *usepitchenv;
 
@@ -275,6 +295,9 @@ class FormantDlg : public wxPanel
 		wxStaticText *t_orig_frame;
 		wxStaticText *t_orig_seq;
 
+		wxSpinCtrl	*s_klatt[N_KLATTP];
+		wxStaticText *t_klatt[N_KLATTP];
+
 	private:
 		wxStaticText *t_lab[10];
       wxStaticText *t_labpk[N_PEAKS];
@@ -346,7 +369,7 @@ extern void WavegenInit(int samplerate, int wavemult_fact);
 extern void WavegenInitPkData(int);  // initialise envelope data
 extern void SetPitch(int length, unsigned char *env, int pitch1, int pitch2);
 extern void SetSynthHtab(int length_mS, USHORT *ht1, int nh1, float pitch1, USHORT *ht2, int nh2, float pitch2);
-extern void MakeWaveFile();
+extern void MakeWaveFile(int synthesis_method);
 extern void MakeWaveBuf(int length, UCHAR *buf);
 
 extern int OpenWaveFile(const char *path, int samplerate);

src/spectdisplay.cpp

@@ -244,10 +244,19 @@ void SpectDisplay::ReadDialogValues()
 void SpectDisplay::RefreshDialogValues(int type)
 //{=============================================
 {
+	int ix;
+	SpectFrame *sf;
+
+	sf = spectseq->frames[sframe];
+
 	spectseq->amplitude = formantdlg->t_amplitude->GetValue();
-	spectseq->frames[sframe]->amp_adjust = formantdlg->t_ampframe->GetValue();
-	spectseq->frames[sframe]->length_adjust =
-		formantdlg->t_timeframe->GetValue() - spectseq->GetFrameLength(sframe,0,NULL);
+	sf->amp_adjust = formantdlg->t_ampframe->GetValue();
+	sf->length_adjust = formantdlg->t_timeframe->GetValue() - spectseq->GetFrameLength(sframe,0,NULL);
+
+	for(ix=0; ix<9; ix++)
+	{
+		sf->klatt_param[ix] = formantdlg->s_klatt[ix]->GetValue();
+	}
 
 	if(type==0)
 	{
@@ -472,7 +481,7 @@ void SpectDisplay::OnMenu(wxCommandEvent& event)
 	int code;
 	wxKeyEvent keyevent;
 	SpectFrame *sf;
-	static int key[] = {0x1044,0x3044,0x104e,0x1047,0x105a,0x1051,0x3051,WXK_F1,WXK_F2,0x1049,WXK_F10};
+	static int key[] = {0x1044,0x3044,0x104e,0x1047,0x105a,0x1051,0x3051,WXK_F1,WXK_F2,0x1049,WXK_F10,0x104b};
 
 	id = event.GetId();
 
@@ -624,13 +633,19 @@ void SpectDisplay::OnKey(wxKeyEvent& event)
 		if(event.ControlDown())
 		{
 			// CTRL, rotate, make right slope steeper
-			pk->pkright-= 5;
+			if(spectseq->synthesizer_type==0)
+				pk->pkright-= 5;
 			pk->pkwidth += 5;
 		}
 		else
 		{
-			pk->pkright -= 10;
+			if(spectseq->synthesizer_type==0)
+				pk->pkright -= 10;
 			pk->pkwidth -= 10;
+			if(pk->pkright < 0)
+				pk->pkright = 0;
+			if(pk->pkwidth < 0)
+				pk->pkwidth = 0;
 		}
 		field = 4;
 		display = 1;
@@ -640,12 +655,14 @@ void SpectDisplay::OnKey(wxKeyEvent& event)
 		if(event.ControlDown())
 		{
 			// CTRL: rotate, make left slope steeper
-			pk->pkright += 5;
+			if(spectseq->synthesizer_type==0)
+				pk->pkright += 5;
 			pk->pkwidth -= 5;
 		}
 		else
 		{
-			pk->pkright += 10;
+			if(spectseq->synthesizer_type==0)
+				pk->pkright += 10;
 			pk->pkwidth += 10;
 		}
 		field = 4;
@@ -653,19 +670,37 @@ void SpectDisplay::OnKey(wxKeyEvent& event)
 		break;
 
 	case '<':   // width--
-		pk->pkright -= 2;
+		if(spectseq->synthesizer_type==0)
+			pk->pkright -= 2;
 		pk->pkwidth -= 2;
+		if(pk->pkwidth < 0)
+			pk->pkwidth = 0;
 		display = 1;
 		field = 4;
 		break;
 
 	case '>':   // width++
-		pk->pkright += 2;
+		if(spectseq->synthesizer_type==0)
+			pk->pkright += 2;
 		pk->pkwidth += 2;
 		display = 1;
 		field = 4;
 		break;
 
+	case '[':   // width--
+		pk->pkright -= 1;
+		if(pk->pkright < 0)
+			pk->pkright = 0;
+		display = 1;
+		field = 4;
+		break;
+
+	case ']':   // width++
+		pk->pkright += 1;
+		display = 1;
+		field = 4;
+		break;
+
 	case '/':  // make left=right
 		i = pk->pkwidth + pk->pkright;
 		pk->pkwidth = pk->pkright = i/2;
@@ -708,6 +743,11 @@ void SpectDisplay::OnKey(wxKeyEvent& event)
 		display = 2;
 		break;
 
+	case 0x104b:   // CTRL-K
+		spectseq->SetKlattDefaults();
+		display = 3;
+		break;
+
 	case 0x104d:   // CTRL-M
 		sf->ToggleMarker(1);
       display = 2;
@@ -825,25 +865,28 @@ void SpectDisplay::OnKey(wxKeyEvent& event)
 		sf = spectseq->frames[sframe];
 		pk = &sf->peaks[pk_num];
 
-		if(pk->pkwidth < 50) pk->pkwidth = 50; // min. width
-		if(pk->pkright < 50) pk->pkright = 50;
+		if(spectseq->synthesizer_type==0)
+		{
+			if(pk->pkwidth < 50) pk->pkwidth = 50; // min. width
+			if(pk->pkright < 50) pk->pkright = 50;
 
-		// ensure minimum separation between peaks & prevent crossover
-		if(direction > 0)
-		{
-			for(i=pk_num+1; i<N_PEAKS; i++)
+			// ensure minimum separation between peaks & prevent crossover
+			if(direction > 0)
 			{
-				if(sf->peaks[i].pkfreq < sf->peaks[i-1].pkfreq + 100)
-					sf->peaks[i].pkfreq = sf->peaks[i-1].pkfreq + 100;
+				for(i=pk_num+1; i<N_PEAKS; i++)
+				{
+					if(sf->peaks[i].pkfreq < sf->peaks[i-1].pkfreq + 100)
+						sf->peaks[i].pkfreq = sf->peaks[i-1].pkfreq + 100;
+				}
 			}
-		}
-		else
-		if(direction < 0)
-		{
-			for(i=pk_num-1; i>=0; i--)
+			else
+			if(direction < 0)
 			{
-				if(sf->peaks[i].pkfreq > sf->peaks[i+1].pkfreq - 100)
-					sf->peaks[i].pkfreq = sf->peaks[i+1].pkfreq - 100;
+				for(i=pk_num-1; i>=0; i--)
+				{
+					if(sf->peaks[i].pkfreq > sf->peaks[i+1].pkfreq - 100)
+						sf->peaks[i].pkfreq = sf->peaks[i+1].pkfreq - 100;
+				}
 			}
 		}
 
@@ -1008,6 +1051,7 @@ void InitSpectrumDisplay()
 	menu_spectdisplay->Append(202,_T("Copy Peaks Up	SHIFT+CTRL+D"));
 	menu_spectdisplay->Append(204,_T("Grid (toggle)	CTRL+G"));
 	menu_spectdisplay->Append(210,_T("Interpolate (percentage)	CTRL+I"));
+	menu_spectdisplay->Append(212,_T("Convert sequence to Klatt format	CTRL+K"));
 	menu_spectdisplay->Append(203,_T("Marker (toggle)	CTRL+N"));
 	menu_spectdisplay->Append(206,_T("Show Interpolation	CTRL+Q"));
 	menu_spectdisplay->Append(207,_T("Hide Interpolation	CTRL+SHIFT+Q"));

src/spectseq.cpp

@@ -34,8 +34,12 @@
 
 #define MAX_HARMONIC  400           // 400 * 50Hz = 20 kHz, more than enough
 
-int  SpeakNextClause(FILE *f_text, const void *text_in, int control);
-extern void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2);
+#define KLATT_TEST
+
+extern void SetSynth_Klatt(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *v, int control);
+extern int Wavegen_Klatt(int resume);
+
+extern void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *v);
 extern int Wavegen();
 extern void CloseWaveFile2();
 extern FILE *f_wave;
@@ -100,18 +104,27 @@ static void PeaksZero(peak_t *sp, peak_t *zero)
 
 
 
-void MakeWaveFile()
-{//================
+void MakeWaveFile(int synthesis_method)
+{//====================================
 	int result=1;
+	int resume=0;
 	unsigned char wav_outbuf[1024];
 
 	while(result != 0)
 	{
 		out_ptr = out_start = wav_outbuf;
 		out_end = &wav_outbuf[sizeof(wav_outbuf)];
-		result = Wavegen();
+
+#ifdef KLATT_TEST
+		if(synthesis_method == 1)
+			result = Wavegen_Klatt(resume);
+		else
+#endif
+			result = Wavegen();
+
 		if(f_wave != NULL)
 			fwrite(wav_outbuf, 1, out_ptr - wav_outbuf, f_wave);
+		resume=1;
 	}
 }  // end of MakeWaveFile
 
@@ -135,6 +148,7 @@ SpectSeq::SpectSeq(int n)
 
 	max_x = 3000;
 	max_y = 1;
+	synthesizer_type = 0;
 }
 
 SpectSeq::~SpectSeq()
@@ -378,7 +392,7 @@ void SpectSeq::Load2(wxInputStream& stream, int import, int n)
 		}
 		else
 		{
-			if(frame->Load(stream) != 0) break;
+			if(frame->Load(stream, synthesizer_type) != 0) break;
 		}
 
 		frames[numframes++] = frame;
@@ -493,8 +507,19 @@ int SpectSeq::Load(wxInputStream & stream)
 		stream.SeekI(4);
 		return(ImportSPC2(stream));
 	}
-	if(id1 != FILEID1_SPECTSEQ || id2 != FILEID2_SPECTSEQ)
+	else
+	if((id1 == FILEID1_SPECTSEQ) && (id2 == FILEID2_SPECTSEQ))
 	{
+			synthesizer_type = 0;   // eSpeak formants
+	}
+	else
+	if((id1 == FILEID1_SPECTSEQ) && (id2 == FILEID2_SPECTSEK))
+	{
+			synthesizer_type = 1;   // formants for Klatt synthesizer
+	}
+	else
+	{
+		// Praat analysis data
 		stream.SeekI(0);
 		return(Import(stream));
 	}
@@ -530,7 +555,10 @@ int SpectSeq::Save(wxOutputStream &stream, int selection)
 	wxDataOutputStream s(stream);
 
 	s.Write32(FILEID1_SPECTSEQ);
-	s.Write32(FILEID2_SPECTSEQ);
+	if(synthesizer_type == 1)
+		s.Write32(FILEID2_SPECTSEK);
+	else
+		s.Write32(FILEID2_SPECTSEQ);
 
 	s.WriteString(name);
 	s.Write16(count);
@@ -549,6 +577,23 @@ int SpectSeq::Save(wxOutputStream &stream, int selection)
 
 
 
+void SpectSeq::SetKlattDefaults(void)
+{//==================================
+	int ix;
+
+	synthesizer_type = 1;
+
+	for(ix=0; ix<numframes; ix++)
+	{
+		if(frames[ix]->keyframe)
+		{
+			frames[ix]->KlattDefaults();
+		}
+		frames[ix]->synthesizer_type = 1;
+	}
+}
+
+
 
 void SpectSeq::ConstructVowel(void)
 {//================================
@@ -785,6 +830,7 @@ void SpectSeq::CopyDown(int frame, int direction)
 				frames[frame]->peaks[pk].pkwidth = frames[f1]->peaks[pk].pkwidth;
 				frames[frame]->peaks[pk].pkright = frames[f1]->peaks[pk].pkright;
 			}
+			memcpy(frames[frame]->klatt_param, frames[f1]->klatt_param, sizeof(frames[frame]->klatt_param));
 			break;
 		}
 	}
@@ -878,46 +924,88 @@ void SpectSeq::ApplyAmp_adjust(SpectFrame *sp, peak_t *peaks)
 	int  y;
 
 	memcpy(peaks,sp->peaks,sizeof(*peaks)*N_PEAKS);
-	for(ix=0; ix<N_PEAKS; ix++)
+
+	if(synthesizer_type == 0)
 	{
-		y = peaks[ix].pkheight * sp->amp_adjust * amplitude;
-		peaks[ix].pkheight = y / 10000;
+		for(ix=0; ix<N_PEAKS; ix++)
+		{
+			y = peaks[ix].pkheight * sp->amp_adjust * amplitude;
+			peaks[ix].pkheight = y / 10000;
+		}
 	}
 }  // end of ApplyAmp_adjust
 
 
 
-void PeaksToFrame(peak_t *pks, frame_t *fr)
-{//========================================
+void PeaksToFrame(SpectFrame *sp1, peak_t *pks, frame_t *fr)
+{//=========================================================
 	int  ix;
-	int  x;
+	int  x, x2;
 
 	for(ix=0; ix<N_PEAKS; ix++)
 	{
 		fr->ffreq[ix] = pks[ix].pkfreq;
-		fr->fheight[ix] = pks[ix].pkheight >> 6;
+
+		if(sp1->synthesizer_type==0)
+		{
+			x = pks[ix].pkheight >> 6;
+		}
+		else
+		{
+			x = pks[ix].pkheight >> 7;
+			fr->fwidth6 = pks[6].pkwidth >> 1;
+			fr->fright6 = pks[6].pkright;
+		}
+
+		if(x > 255)
+			x = 255;
+		fr->fheight[ix] = x;
+
 		if(ix < 6)
 		{
-			if((x = (pks[ix].pkwidth >> 2)) > 255)
+			x = pks[ix].pkwidth/2;
+			x2 = pks[ix].pkright;
+
+			if(sp1->synthesizer_type == 0)
+			{
+				x /= 2;
+				x2 /= 4;
+			}
+			if(x > 255)
 				x = 255;
 			fr->fwidth[ix] = x;
 
-			if((x = (pks[ix].pkright >> 2)) > 255)
-				x = 255;
-			fr->fright[ix] = x;
+			if(x2 > 255)
+				x2 = 255;
+			fr->fright[ix] = x2;
 		}
 	}
+
+	for(ix=0; ix<N_KLATTP; ix++)
+	{
+		fr->klattp[ix] = sp1->klatt_param[ix];
+	}
 }
 
-static void SetSynth_mS(int length_mS, peak_t *sp1, peak_t *sp2)
-{//=============================================================
+static void SetSynth_mS(int length_mS, SpectFrame *sp1, SpectFrame *sp2, peak_t *pks1, peak_t *pks2, int control)
+{//==============================================================================================================
 	static frame_t fr1, fr2;
 
-	PeaksToFrame(sp1,&fr1);
-	PeaksToFrame(sp2,&fr2);
+	PeaksToFrame(sp1,pks1,&fr1);
+	PeaksToFrame(sp2,pks2,&fr2);
+
+#ifdef KLATT_TEST
+	if(sp1->synthesizer_type == 1)
+	{
+		SetSynth_Klatt((length_mS * samplerate) / 1000, 0, &fr1, &fr2, voice, control);    // convert mS to samples
+	}
+	else
+#endif
+	{
+		SetSynth((length_mS * samplerate) / 1000, 0, &fr1, &fr2, voice);    // convert mS to samples
+	}
+};
 
-	SetSynth((length_mS * samplerate) / 1000, 0, &fr1, &fr2);    // convert mS to samples
-}
 
 
 
@@ -939,6 +1027,7 @@ void SpectSeq::MakeWave(int start, int end, PitchEnvelope &pitch)
 	peak_t peaks1[N_PEAKS];
 	peak_t peaks2[N_PEAKS];
 
+KlattInit();
 	SpeakNextClause(NULL,NULL,2);  // stop speaking file
 
 	if(numframes==0) return;
@@ -1007,23 +1096,39 @@ void SpectSeq::MakeWave(int start, int end, PitchEnvelope &pitch)
 			{
 				ApplyAmp_adjust(sp1,peaks1);
 				ApplyAmp_adjust(sp2,peaks2);
+
 				if(first)
 				{
-					PeaksZero(peaks1,peaks0);  // fade in
-					SetSynth_mS(20,peaks0,peaks1);
-					MakeWaveFile();
+					if(synthesizer_type == 1)
+					{
+						memcpy(peaks0,peaks1,sizeof(peaks0));
+					}
+					else
+					{
+						PeaksZero(peaks1,peaks0);  // fade in
+					}
+					SetSynth_mS(20,sp1,sp1,peaks0,peaks1,0);
+					MakeWaveFile(synthesizer_type);
 					first=0;
 				}
 
 				length = int(sp1->length * lfactor);
-				SetSynth_mS(length,peaks1,peaks2);
-				MakeWaveFile();
+				SetSynth_mS(length,sp1,sp2,peaks1,peaks2,0);
+				MakeWaveFile(synthesizer_type);
 			}
 		}
 	}
-	PeaksZero(peaks2,peaks0);  // fade out
-	SetSynth_mS(30,peaks2,peaks0);
-	MakeWaveFile();
+
+	if(synthesizer_type == 1)
+	{
+		memcpy(peaks0,peaks2,sizeof(peaks0));
+	}
+	else
+	{
+		PeaksZero(peaks2,peaks0);  // fade out
+	}
+	SetSynth_mS(30,sp2,sp2,peaks2,peaks0,2);
+	MakeWaveFile(synthesizer_type);
 
 	CloseWaveFile2();
 	PlayWavFile(fname_speech);
@@ -1054,6 +1159,7 @@ void SpectFrame::MakeWave(int control, PitchEnvelope &pitche, int amplitude, int
 	char *fname_speech;
 //	USHORT htab0[600];
 
+KlattInit();
 	SpeakNextClause(NULL,NULL,2);  // stop speaking file
 
 	length = duration;
@@ -1069,23 +1175,31 @@ void SpectFrame::MakeWave(int control, PitchEnvelope &pitche, int amplitude, int
 	if(OpenWaveFile2(fname_speech) != 0)
 		return;
 
+
 	if(control==0)
 	{
 		memcpy(peaks1,peaks,sizeof(peaks1));
-		for(ix=0; ix<N_PEAKS; ix++)
-		{
-			y = peaks1[ix].pkheight * amp_adjust * amplitude;
-			peaks1[ix].pkheight = y/10000;
-		}
-			
-		PeaksZero(peaks1,peaks0);
 
-		SetSynth_mS(20,peaks0,peaks1);
-		MakeWaveFile();
-		SetSynth_mS(length,peaks1,peaks1);
-		MakeWaveFile();
-		SetSynth_mS(30,peaks1,peaks0);
-		MakeWaveFile();
+		if(synthesizer_type == 0)
+		{
+			for(ix=0; ix<N_PEAKS; ix++)
+			{
+				y = peaks1[ix].pkheight * amp_adjust * amplitude;
+				peaks1[ix].pkheight = y/10000;
+			}
+			PeaksZero(peaks1,peaks0);
+		}
+		else
+		{
+			memcpy(peaks0,peaks1,sizeof(peaks0));
+		}
+	
+		SetSynth_mS(20,this,this,peaks0,peaks1,0);
+		MakeWaveFile(synthesizer_type);
+		SetSynth_mS(length,this,this,peaks1,peaks1,0);
+		MakeWaveFile(synthesizer_type);
+		SetSynth_mS(30,this,this,peaks1,peaks0,2);
+		MakeWaveFile(synthesizer_type);
 	}
 	else
 	{

src/speech.h

@@ -21,6 +21,7 @@
 #include <sys/types.h>
 
 // conditional compilation options
+#define INCLUDE_KLATT
 
 #if defined(BYTE_ORDER) && BYTE_ORDER == BIG_ENDIAN
 #define ARCH_BIG

src/synth_mbrola.cpp

@@ -77,7 +77,7 @@ BOOL load_MBR()
 	if(hinstDllMBR != NULL)
 		return TRUE;   // already loaded 
 
-	if (!(hinstDllMBR=LoadLibrary("mbrola.dll")))
+	if (!(hinstDllMBR=LoadLibraryA("mbrola.dll")))
 		return FALSE;
 	init_MBR			=(PROCIC) GetProcAddress(hinstDllMBR,"init_MBR");
 	write_MBR			=(PROCIC) GetProcAddress(hinstDllMBR,"write_MBR");
@@ -179,7 +179,7 @@ espeak_ERROR LoadMbrolaTable(const char *mbrola_voice, const char *phtrans, int
 
 #ifdef USE_MBROLA_LIB
 #ifdef PLATFORM_WINDOWS
-	setVolumeRatio_MBR((float)(mbrola_control & 0xff) /16.0);
+	setVolumeRatio_MBR((float)(mbrola_control & 0xff) /16.0f);
 #else
 	mbrolib_get_parameter(mb_handle,&m_parameters);
 	m_parameters.ignore_error = 1;

src/synthdata.cpp

@@ -35,7 +35,7 @@
 #include "translate.h"
 #include "wave.h"
 
-const char *version_string = "1.39.26  15.Nov.08";
+const char *version_string = "1.39.28  22.Nov.08";
 const int version_phdata  = 0x013900;
 
 int option_device_number = -1;
@@ -375,9 +375,10 @@ frameref_t *LookupSpect(PHONEME_TAB *this_ph, PHONEME_TAB *prev_ph, PHONEME_TAB
 	int  length1;
 	int  length_std;
 	int  length_factor;
-	SPECT_SEQ *seq;
-	SPECT_SEQ *seq2;
+	SPECT_SEQ *seq, *seq2;
+	SPECT_SEQK *seqk, *seqk2;
 	PHONEME_TAB *next2_ph;
+	frame_t *frame;
 	static frameref_t frames_buf[N_SEQ_FRAMES];
 	
 	PHONEME_TAB *other_ph;
@@ -389,6 +390,7 @@ frameref_t *LookupSpect(PHONEME_TAB *this_ph, PHONEME_TAB *prev_ph, PHONEME_TAB
 	if((ix = LookupSound(this_ph,other_ph,which,match_level,0)) < 4)
 		return(NULL);
 	seq = (SPECT_SEQ *)(&spects_data[ix]);
+	seqk = (SPECT_SEQK *)seq;
 	nf = seq->n_frames;
 
 
@@ -397,15 +399,20 @@ frameref_t *LookupSpect(PHONEME_TAB *this_ph, PHONEME_TAB *prev_ph, PHONEME_TAB
 
 	seq_break = 0;
 	length1 = 0;
+
 	for(ix=0; ix<nf; ix++)
 	{
-		frames_buf[ix].frame = &seq->frame[ix];
-		frames_buf[ix].frflags = seq->frame[ix].frflags;
-		frames_buf[ix].length = seq->frame[ix].length;
-		if(seq->frame[ix].frflags & FRFLAG_VOWEL_CENTRE)
+		if(seq->frame[0].frflags & FRFLAG_KLATT)
+			frame = &seqk->frame[ix];
+		else
+			frame = (frame_t *)&seq->frame[ix];
+		frames_buf[ix].frame = frame;
+		frames_buf[ix].frflags = frame->frflags;
+		frames_buf[ix].length = frame->length;
+		if(frame->frflags & FRFLAG_VOWEL_CENTRE)
 			seq_break = ix;
 	}
-	
+
 	frames = &frames_buf[0];
 	if(seq_break > 0)
 	{
@@ -462,14 +469,23 @@ frameref_t *LookupSpect(PHONEME_TAB *this_ph, PHONEME_TAB *prev_ph, PHONEME_TAB
 		// a secondary reference has been returned, which is not a wavefile
 		// add these spectra to the main sequence
 		seq2 = (SPECT_SEQ *)(&spects_data[wavefile_ix]);
+		seqk2 = (SPECT_SEQK *)seq2;
 	
 		// first frame of the addition just sets the length of the last frame of the main seq
 		nf--;
 		for(ix=0; ix<seq2->n_frames; ix++)
 		{
-			frames[nf].length = seq2->frame[ix].length;
+			if(seq2->frame[0].frflags & FRFLAG_KLATT)
+				frame = &seqk2->frame[ix];
+			else
+				frame = (frame_t *)&seq2->frame[ix];
+
+			frames[nf].length = frame->length;
 			if(ix > 0)
-				frames[nf].frame = &seq2->frame[ix];
+			{
+				frames[nf].frame = frame;
+				frames[nf].frflags = frame->frflags;
+			}
 			nf++;
 		}
 		wavefile_ix = 0;

src/synthesize.cpp

@@ -354,6 +354,15 @@ static void set_frame_rms(frame_t *fr, int new_rms)
 	849,851,853,856,858,861,863,865,868,870,872,875,877,879,882,884,
 	886,889,891,893,896,898,900,902};
 
+	if(fr->frflags & FRFLAG_KLATT)
+	{
+		if(new_rms == -1)
+		{
+			fr->klattp[KLATT_AV] = 50;
+		}
+		return;
+	}
+ 
 	if(fr->rms == 0) return;    // check for divide by zero
 	x = (new_rms * 64)/fr->rms;
 	if(x >= 200) x = 199;
@@ -375,6 +384,9 @@ static void formants_reduce_hf(frame_t *fr, int level)
 	int  ix;
 	int  x;
 
+	if(fr->frflags & FRFLAG_KLATT)
+		return;
+ 
 	for(ix=2; ix<N_PEAKS; ix++)
 	{
 		x = fr->fheight[ix] * level;
@@ -540,6 +552,10 @@ static short vcolouring[N_VCOLOUR][5] = {
 
 		next_rms = seq[1].frame->rms;
 
+if(fr->frflags & FRFLAG_KLATT)
+{
+	fr->klattp[KLATT_AV] = 53;   // reduce the amplituide of the start of a vowel
+}
 		if(f2 != 0)
 		{
 			if(rms & 0x20)
@@ -661,6 +677,9 @@ static void SmoothSpect(void)
 	q = wcmdq[syllable_centre];
 	frame_centre = (frame_t *)q[2];
 
+//if(frame_centre->frflags & FRFLAG_KLATT)
+//	return;  // TESTING
+
 	// backwards
 	ix = syllable_centre -1;
 	frame = frame2 = frame_centre;
@@ -672,7 +691,7 @@ static void SmoothSpect(void)
 		if(q[0] == WCMD_PAUSE || q[0] == WCMD_WAVE)
 			break;
 
-		if(q[0] == WCMD_SPECT || q[0] == WCMD_SPECT2)
+		if(q[0] <= WCMD_SPECT2)
 		{
 			len = q[1] & 0xffff;
 
@@ -759,7 +778,7 @@ static void SmoothSpect(void)
 		if(q[0] == WCMD_PAUSE || q[0] == WCMD_WAVE)
 			break;
 
-		if(q[0] == WCMD_SPECT || q[0] == WCMD_SPECT2)
+		if(q[0] <= WCMD_SPECT2)
 		{
 
 			len = q[1] & 0xffff;
@@ -834,7 +853,7 @@ static void SmoothSpect(void)
 	}
 
 	syllable_start = syllable_end;
-}
+}  //  end of SmoothSpect
 
 
 static void StartSyllable(void)
@@ -891,22 +910,29 @@ if(which==1)
 	if(frames == NULL)
 		return(0);   // not found
 
+	frame1 = frames[0].frame;
+	frame1_length = frames[0].length;
+	if(frame1->frflags & FRFLAG_KLATT)
+		wcmd_spect = WCMD_KLATT;
+
 	if(wavefile_ix == 0)
 	{
 		if(wave_flag)
 		{
 			// cancel any wavefile that was playing previously
 			wcmd_spect = WCMD_SPECT2;
+			if(frame1->frflags & FRFLAG_KLATT)
+				wcmd_spect = WCMD_KLATT2;
 			wave_flag = 0;
 		}
 		else
 		{
 			wcmd_spect = WCMD_SPECT;
+			if(frame1->frflags & FRFLAG_KLATT)
+				wcmd_spect = WCMD_KLATT;
 		}
 	}
 
-	frame1 = frames[0].frame;
-	frame1_length = frames[0].length;
 	if(last_frame != NULL)
 	{
 		if(((last_frame->length < 2) || (last_frame->frflags & FRFLAG_VOWEL_CENTRE))

src/synthesize.h

@@ -22,11 +22,13 @@
 
 #define MAX_HARMONIC  400           // 400 * 50Hz = 20 kHz, more than enough
 #define N_SEQ_FRAMES   25           // max frames in a spectrum sequence (real max is ablut 8)
+#define STEPSIZE  64                // 2.9mS at 22 kHz sample rate
 
 #define    PITCHfall   0
 #define    PITCHrise   1
 
 // flags set for frames within a spectrum sequence
+#define FRFLAG_KLATT           0x01   // this frame includes extra data for Klatt synthesizer
 #define FRFLAG_VOWEL_CENTRE    0x02   // centre point of vowel
 #define FRFLAG_LEN_MOD         0x04   // reduce effect of length adjustment
 #define FRFLAG_BREAK_LF        0x08   // but keep f3 upwards
@@ -75,6 +77,18 @@ typedef struct {
    short pkright;
 }  peak_t;
 
+#define N_KLATTP   10   // this affects the phoneme data file format
+#define KLATT_AV      0
+#define KLATT_Kopen   1
+#define KLATT_Skew    2
+#define KLATT_Tilt    3
+#define KLATT_Turb    4
+#define KLATT_Aspr    5
+#define KLATT_AVp     6  // this is after the parameters which can be change by the Voice
+#define KLATT_Fric    7
+#define KLATT_FricBP  8
+#define KLATT_spare1  9
+
 typedef struct {
 	short frflags;
 	unsigned char length;
@@ -83,8 +97,20 @@ typedef struct {
 	unsigned char fheight[9];
 	unsigned char fwidth[6];          // width/4
 	unsigned char fright[6];          // width/4
+	unsigned char fwidth6, fright6;
+	unsigned char klattp[N_KLATTP];
 } frame_t;
 
+typedef struct {
+	short frflags;
+	unsigned char length;
+	unsigned char rms;
+	short ffreq[9];
+	unsigned char fheight[9];
+	unsigned char fwidth[6];          // width/4
+	unsigned char fright[6];          // width/4
+} frame_t2;   //  the original, without Klatt additions, used for file "phondata" 
+
 
 
 // formant data used by wavegen
@@ -103,6 +129,26 @@ typedef struct {
 	DOUBLEX right_inc;
 }  wavegen_peaks_t;
 
+typedef struct {
+unsigned char *pitch_env;
+int pitch;          // pitch Hz*256
+int pitch_ix;       // index into pitch envelope (*256)
+int pitch_inc;      // increment to pitch_ix
+int pitch_base;     // Hz*256 low, before modified by envelope
+int pitch_range;    // Hz*256 range of envelope
+
+unsigned char *mix_wavefile;  // wave file to be added to synthesis
+int n_mix_wavefile;       // length in bytes
+int mix_wave_scale;         // 0=2 byte samples
+int mix_wave_amp;
+int mix_wavefile_ix;
+
+int amplitude;
+int amplitude_v;
+int prev_was_synth;  // previous sound was synthesized (not a played wave or pause)
+} WGEN_DATA;
+
+
 typedef struct {
 	double a;
 	double b;
@@ -113,11 +159,19 @@ typedef struct {
 
 
 typedef struct {
-   short length;
+   short length_total;  // not used
+   unsigned char  n_frames;
+   unsigned char  flags;
+   frame_t2  frame[N_SEQ_FRAMES];     // max. frames in a spectrum sequence
+} SPECT_SEQ;   // sequence of espeak formant frames
+
+typedef struct {
+   short length_total;  // not used
    unsigned char  n_frames;
    unsigned char  flags;
    frame_t  frame[N_SEQ_FRAMES];     // max. frames in a spectrum sequence
-} SPECT_SEQ;
+} SPECT_SEQK;   // sequence of klatt formants frames
+
 
 typedef struct {
 	short length;
@@ -176,16 +230,18 @@ extern unsigned char env_frise[128];
 extern unsigned char pitch_adjust_tab[MAX_PITCH_VALUE+1];
 
 // queue of commands for wavegen
-#define WCMD_AMPLITUDE 1
-#define WCMD_PITCH	2
+#define WCMD_KLATT	1
+#define WCMD_KLATT2	2
 #define WCMD_SPECT	3
 #define WCMD_SPECT2	4
 #define WCMD_PAUSE	5
 #define WCMD_WAVE    6
 #define WCMD_WAVE2   7
-#define WCMD_MARKER	8
-#define WCMD_VOICE   9
-#define WCMD_EMBEDDED 10
+#define WCMD_AMPLITUDE 8
+#define WCMD_PITCH	9
+#define WCMD_MARKER	10
+#define WCMD_VOICE   11
+#define WCMD_EMBEDDED 12
 
 
 #define N_WCMDQ   160
@@ -278,3 +334,6 @@ int PauseLength(int pause, int control);
 int LookupPhonemeTable(const char *name);
 
 void InitBreath(void);
+
+void KlattInit();
+int Wavegen_Klatt2(int length, int modulation, int resume, frame_t *fr1, frame_t *fr2);

src/tr_languages.cpp

@@ -515,6 +515,7 @@ SetLengthMods(tr,3);  // all equal
 			tr->langopts.param[LOPT_SONORANT_MIN] = 130;  // limit the shortening of sonorants before short vowels
 			tr->langopts.param[LOPT_REDUCE] = 1;        // reduce vowels even if phonemes are specified in it_list
 			tr->langopts.numbers = 0x2709 + NUM_ROMAN;
+			tr->langopts.accents = 2;   // Say "Capital" after the letter.
 		}
 		break;
 
@@ -769,8 +770,9 @@ SetLengthMods(tr,3);  // all equal
 			static const unsigned char stress_amps_sq[8] = {17,12, 17,17, 20,20, 24,22 };
 			tr = new Translator();
 			SetupTranslator(tr,stress_lengths_sq,stress_amps_sq);
-			tr->langopts.stress_flags =  0x4; 
+			tr->langopts.stress_flags =  0x4 + 0x10; 
 			SetLetterVowel(tr,'y');
+			tr->langopts.numbers = 0x69 + 0x8000;
 		}
 		break;
 

src/translate.cpp

@@ -1747,7 +1747,11 @@ int Translator::TranslateWord2(char *word, WORD_TAB *wtab, int pre_pause, int ne
 			first_phoneme = 0;
 		}
 	}
-	plist2->sourceix = source_ix;  // ????
+	// don't set new-word if there is a hyphen before it
+	if((word_flags & FLAG_HYPHEN) == 0)
+	{
+		plist2->sourceix = source_ix;
+	}
 
 	end_stressed_vowel = 0;
 	if((stress >= 4) && (phoneme_tab[ph_list2[n_ph_list2-1].phcode]->type == phVOWEL))

src/translate.h

@@ -29,7 +29,7 @@
 #define N_RULE_GROUP2    120          // max num of two-letter rule chains
 #define N_HASH_DICT     1024
 #define N_CHARSETS        20
-#define N_LETTER_GROUPS   25
+#define N_LETTER_GROUPS   26
 
 
 /* dictionary flags, word 1 */

src/transldlg.cpp

@@ -249,6 +249,7 @@ void MakeWave2(PHONEME_LIST *p, int n_phonemes)
 	char *fname_speech;
 	int n_ph;
 
+	KlattInit();
 	n_ph = n_phonemes;
 	fname_speech = WavFileName();
 	OpenWaveFile2(fname_speech);
@@ -275,9 +276,9 @@ TranslDlg::TranslDlg(wxWindow *parent) : wxPanel(parent)
 {//=====================================================
 
 	t_source = new wxTextCtrl(this,T_SOURCE,_T(""),wxPoint(0,4),
-		wxSize(200,250),wxTE_MULTILINE);
+		wxSize(204,250),wxTE_MULTILINE);
 	t_phonetic = new wxTextCtrl(this,T_PHONETIC,_T(""),wxPoint(0,262),
-		wxSize(200,350),wxTE_MULTILINE);
+		wxSize(204,350),wxTE_MULTILINE);
 
 	t_translate = new wxButton(this,T_TRANSLATE,_T("Translate"),wxPoint(4,620));
 	t_translate = new wxButton(this,T_RULES,_T("Show Rules"),wxPoint(4,652));

src/voice.h

@@ -40,6 +40,7 @@ typedef struct {
 	int formant_factor;      // adjust nominal formant frequencies by this  because of the voice's pitch (256ths)
 	int consonant_amp;     // amplitude of unvoiced consonants
 	int consonant_ampv;    // amplitude of the noise component of voiced consonants
+	int klatt[8];
 
 	// parameters used by Wavegen
 	short freq[N_PEAKS];    // 100% = 256
@@ -68,7 +69,7 @@ extern USHORT voice_pcnt[N_PEAKS+1][3];
 
 
 extern voice_t *voice;
-extern int tone_points[10];
+extern int tone_points[12];
 
 const char *SelectVoice(espeak_VOICE *voice_select);
 espeak_VOICE *SelectVoiceByName(espeak_VOICE **voices, const char *name);

src/voices.cpp

@@ -49,8 +49,8 @@ MNEM_TAB genders [] = {
 	{"female", 2},
 	{NULL, 0 }};
 
-//int tone_points[10] = {250,140, 1200,110, -1,0, -1,0, -1,0};
-int tone_points[10] = {600,170, 1200,135, 2000,110, 3000,110, -1,0};
+int tone_points[12] = {600,170, 1200,135, 2000,110, 3000,110, -1,0};
+//int tone_points[12] = {250,200,  400,170, 600,170, 1200,135, 2000,110, -1,0};
 
 // limit the rate of change for each formant number
 //static int formant_rate_22050[9] = {50, 104, 165, 230, 220, 220, 220, 220, 220};  // values for 22kHz sample rate
@@ -101,11 +101,13 @@ espeak_VOICE voice_selected;
 #define V_NUMBERS     25
 #define V_OPTION      26
 
-#define V_MBROLA     27
+#define V_MBROLA      27
+#define V_KLATT       28
 
 // these need a phoneme table to have been specified
-#define V_REPLACE    28
-#define V_CONSONANTS 29
+#define V_REPLACE    29
+#define V_CONSONANTS 30
+
 
 
 typedef struct {
@@ -144,6 +146,7 @@ static keywtab_t keyword_tab[] = {
 	{"option",     V_OPTION},
 	{"mbrola",     V_MBROLA},
 	{"consonants", V_CONSONANTS},
+	{"klatt",      V_KLATT},
 
 	// these just set a value in langopts.param[]
 	{"l_dieresis", 0x100+LOPT_DIERESES},
@@ -195,7 +198,7 @@ void SetToneAdjust(voice_t *voice, int *tone_pts)
 	int height2;
 	double rate;
 
-	for(pt=0; pt<10; pt+=2)
+	for(pt=0; pt<12; pt+=2)
 	{
 		if(tone_pts[pt] == -1)
 		{
@@ -225,15 +228,16 @@ void SetToneAdjust(voice_t *voice, int *tone_pts)
 
 void ReadTonePoints(char *string, int *tone_pts)
 {//=============================================
-// tone_pts[] is int[10]
+// tone_pts[] is int[12]
 	int ix;
 
-	for(ix=0; ix<10; ix++)
+	for(ix=0; ix<12; ix++)
 		tone_pts[ix] = -1;
 
-	sscanf(string,"%d %d %d %d %d %d %d %d",
+	sscanf(string,"%d %d %d %d %d %d %d %d %d %d",
 		&tone_pts[0],&tone_pts[1],&tone_pts[2],&tone_pts[3],
-		&tone_pts[4],&tone_pts[5],&tone_pts[6],&tone_pts[7]);
+		&tone_pts[4],&tone_pts[5],&tone_pts[6],&tone_pts[7],
+		&tone_pts[8],&tone_pts[9]);
 }
 
 
@@ -369,10 +373,11 @@ void VoiceReset(int tone_only)
 	voice->echo_amp = 0;
 	voice->flutter = 64;
 	voice->n_harmonic_peaks = 5;
-	voice->peak_shape = 1;
+	voice->peak_shape = 0;
 	voice->voicing = 64;
 	voice->consonant_amp = 100;
 	voice->consonant_ampv = 100;
+	memset(voice->klatt,0,sizeof(voice->klatt));
 
 #ifdef PLATFORM_RISCOS
 	voice->roughness = 1;
@@ -864,6 +869,11 @@ voice_t *LoadVoice(const char *vname, int control)
 			}
 			break;
 
+		case V_KLATT:
+			Read8Numbers(p,voice->klatt);
+			voice->klatt[KLATT_Kopen] -= 40;
+			break;
+
 		default:
 			if((key & 0xff00) == 0x100)
 			{
@@ -1493,13 +1503,12 @@ void GetVoices(const char *path)
 	}
 #else
 #ifdef PLATFORM_WINDOWS
-   WIN32_FIND_DATA FindFileData;
+   WIN32_FIND_DATAA FindFileData;
    HANDLE hFind = INVALID_HANDLE_VALUE;
-   DWORD dwError;
 
 #undef UNICODE         // we need FindFirstFileA() which takes an 8-bit c-string
 	sprintf(fname,"%s\\*",path);
-	hFind = FindFirstFile(fname, &FindFileData);
+	hFind = FindFirstFileA(fname, &FindFileData);
 	if(hFind == INVALID_HANDLE_VALUE)
 		return;
 
@@ -1529,7 +1538,7 @@ void GetVoices(const char *path)
 				voices_list[n_voices_list++] = voice_data;
 			}
 		}
-	} while(FindNextFile(hFind, &FindFileData) != 0);
+	} while(FindNextFileA(hFind, &FindFileData) != 0);
 	FindClose(hFind);
 
 #else

src/vowelchart.cpp

@@ -510,6 +510,7 @@ void MakeVowelLists(void)
 	PHONEME_TAB *ph;
 	FILE *f;
 	SPECT_SEQ *seq;
+	SPECT_SEQK *seqk;
 	frame_t *frame;
 	int match_level;
 	char dirname[sizeof(path_source)+20];
@@ -547,10 +548,20 @@ void MakeVowelLists(void)
 				continue;
 
 			seq = (SPECT_SEQ *)(&spects_data[ix]);
-			frame = &seq->frame[1];
+			seqk = (SPECT_SEQK *)seq;
+
+			if(seq->frame[0].frflags & FRFLAG_KLATT)
+				frame = &seqk->frame[1];
+			else
+				frame = (frame_t *)&seq->frame[1];
+
 			fprintf(f,"%s\t %3d %4d %4d",WordToString(ph->mnemonic),
 					frame->ffreq[1],frame->ffreq[2],frame->ffreq[3]);
-			frame = &seq->frame[seq->n_frames-1];
+
+			if(seq->frame[0].frflags & FRFLAG_KLATT)
+				frame = &seqk->frame[seqk->n_frames-1];
+			else
+				frame = (frame_t *)&seq->frame[seq->n_frames-1];
 			fprintf(f,"   %3d %4d %4d\n",frame->ffreq[1],frame->ffreq[2],frame->ffreq[3]);
 		}
 		fclose(f);

src/wavegen.cpp

@@ -34,6 +34,8 @@
 #include "synthesize.h"
 #include "voice.h"
 
+//#undef INCLUDE_KLATT
+
 #ifdef USE_PORTAUDIO
 #include "portaudio.h"
 #undef USE_PORTAUDIO
@@ -51,10 +53,9 @@
 
 #define PI  3.1415927
 #define PI2 6.283185307
-#define STEPSIZE  64                // 2.9mS at 22 kHz sample rate
 #define N_WAV_BUF   10
 
-static voice_t *wvoice;
+voice_t *wvoice;
 
 FILE *f_log = NULL;
 int option_waveout = 0;
@@ -98,24 +99,13 @@ static int hspect[2][MAX_HARMONIC];         // 2 copies, we interpolate between
 static int max_hval=0;
 
 static int nsamples=0;       // number to do
-static int amplitude = 32;
-static int amplitude_v = 0;
 static int modulation_type = 0;
 static int glottal_flag = 0;
 static int glottal_reduce = 0;
 
-static unsigned char *mix_wavefile = NULL;  // wave file to be added to synthesis
-static int n_mix_wavefile = 0;       // length in bytes
-static int mix_wave_scale = 0;         // 0=2 byte samples
-static int mix_wave_amp = 32;
-static int mix_wavefile_ix = 0;
 
-static int pitch;          // pitch Hz*256
-static int pitch_ix;       // index into pitch envelope (*256)
-static int pitch_inc;      // increment to pitch_ix
-static unsigned char *pitch_env=NULL;
-static int pitch_base;     // Hz*256 low, before modified by envelope
-static int pitch_range;    // Hz*256 range of envelope
+WGEN_DATA wdata;
+
 static int amp_ix;
 static int amp_inc;
 static unsigned char *amplitude_env = NULL;
@@ -385,8 +375,8 @@ static void WcmdqIncHead()
 // data points from which to make the presets for pk_shape1 and pk_shape2
 #define PEAKSHAPEW 256
 static const float pk_shape_x[2][8] = {
-	{0,-0.6, 0.0, 0.6, 1.4, 2.5, 4.5, 5.5},
-	{0,-0.6, 0.0, 0.6, 1.4, 2.0, 4.5, 5.5 }};
+	{0,-0.6f, 0.0f, 0.6f, 1.4f, 2.5f, 4.5f, 5.5f},
+	{0,-0.6f, 0.0f, 0.6f, 1.4f, 2.0f, 4.5f, 5.5f }};
 static const float pk_shape_y[2][8] = {
 	{0,  67,  81,  67,  31,  14,   0,  -6} ,
 	{0,  77,  81,  77,  31,   7,   0,  -6 }};
@@ -737,6 +727,7 @@ void WavegenInit(int rate, int wavemult_fact)
 	if(wavemult_fact == 0)
 		wavemult_fact=60;  // default
 
+	wvoice = NULL;
 	samplerate = samplerate_native = rate;
 	PHASE_INC_FACTOR = 0x8000000 / samplerate;   // assumes pitch is Hz*32
 	Flutter_inc = (64 * samplerate)/rate;
@@ -745,6 +736,9 @@ void WavegenInit(int rate, int wavemult_fact)
 	wavephase = 0x7fffffff;
 	max_hval = 0;
 
+	wdata.amplitude = 32;
+	wdata.prev_was_synth = 0;
+
 	for(ix=0; ix<N_EMBEDDED_VALUES; ix++)
 		embedded_value[ix] = embedded_default[ix];
 
@@ -769,7 +763,11 @@ void WavegenInit(int rate, int wavemult_fact)
 
 	WavegenInitPkData(1);
 	WavegenInitPkData(0);
-	pk_shape = pk_shape2;         // ph_shape2
+	pk_shape = pk_shape2;         // pk_shape2
+
+#ifdef INCLUDE_KLATT
+	KlattInit();
+#endif
 
 #ifdef LOG_FRAMES
 remove("log-espeakedit");
@@ -784,7 +782,7 @@ int GetAmplitude(void)
 	// normal, none, reduced, moderate, strong
 	static const unsigned char amp_emphasis[5] = {16, 16, 10, 16, 22};
 
-	amp = (embedded_value[EMBED_A])*60/100;
+	amp = (embedded_value[EMBED_A])*55/100;
 	general_amplitude = amp * amp_emphasis[embedded_value[EMBED_F]] / 16;
 	return(general_amplitude);
 }
@@ -903,6 +901,13 @@ int PeaksToHarmspect(wavegen_peaks_t *peaks, int pitch, int *htab, int control)
 		}
 	}
 
+// increase bass, up to the F1 peak
+h=1;
+for(f=pitch; f<peaks[1].freq; f+=pitch)
+{
+	htab[h++] += (peaks[1].height * 16);
+}
+
 	// find the nearest harmonic for HF peaks where we don't use shape
 	for(; pk<N_PEAKS; pk++)
 	{
@@ -961,10 +966,10 @@ static void AdvanceParameters()
 	static int Flutter_ix = 0;
 
 	// advance the pitch
-	pitch_ix += pitch_inc;
-	if((ix = pitch_ix>>8) > 127) ix = 127;
-	x = pitch_env[ix] * pitch_range;
-	pitch = (x>>8) + pitch_base;
+	wdata.pitch_ix += wdata.pitch_inc;
+	if((ix = wdata.pitch_ix>>8) > 127) ix = 127;
+	x = wdata.pitch_env[ix] * wdata.pitch_range;
+	wdata.pitch = (x>>8) + wdata.pitch_base;
 
 	amp_ix += amp_inc;
 
@@ -973,9 +978,9 @@ static void AdvanceParameters()
 		Flutter_ix = 0;
 	x = ((int)(Flutter_tab[Flutter_ix >> 6])-0x80) * flutter_amp;
 	Flutter_ix += Flutter_inc;
-	pitch += x;
-	if(pitch < 102400)
-		pitch = 102400;   // min pitch, 25 Hz  (25 << 12)
+	wdata.pitch += x;
+	if(wdata.pitch < 102400)
+		wdata.pitch = 102400;   // min pitch, 25 Hz  (25 << 12)
 
 	if(samplecount == samplecount_start)
 		return;
@@ -1159,26 +1164,26 @@ int Wavegen()
 			{
 				hswitch = 0;
 				harmspect = hspect[0];
-				maxh2 = PeaksToHarmspect(peaks,pitch<<4,hspect[0],0);
+				maxh2 = PeaksToHarmspect(peaks, wdata.pitch<<4, hspect[0], 0);
 
 				// adjust amplitude to compensate for fewer harmonics at higher pitch
-				amplitude2 = (amplitude * pitch)/(100 << 11);
+				amplitude2 = (wdata.amplitude * wdata.pitch)/(100 << 11);
 
             // switch sign of harmonics above about 900Hz, to reduce max peak amplitude
-				h_switch_sign = 890 / (pitch >> 12);
+				h_switch_sign = 890 / (wdata.pitch >> 12);
 			}
 			else
 				AdvanceParameters();
 
 			// pitch is Hz<<12
-			phaseinc = (pitch>>7) * PHASE_INC_FACTOR;
-			cycle_samples = samplerate/(pitch >> 12);  // sr/(pitch*2)
-			hf_factor = pitch >> 11;
+			phaseinc = (wdata.pitch>>7) * PHASE_INC_FACTOR;
+			cycle_samples = samplerate/(wdata.pitch >> 12);  // sr/(pitch*2)
+			hf_factor = wdata.pitch >> 11;
 
 			maxh = maxh2;
 			harmspect = hspect[hswitch];
 			hswitch ^= 1;
-			maxh2 = PeaksToHarmspect(peaks,pitch<<4,hspect[hswitch],1);
+			maxh2 = PeaksToHarmspect(peaks, wdata.pitch<<4, hspect[hswitch], 1);
 
 			SetBreath();
 		}
@@ -1211,11 +1216,11 @@ int Wavegen()
 				for(pk=wvoice->n_harmonic_peaks+1; pk<N_PEAKS; pk++)
 				{
 					// find the nearest harmonic for HF peaks where we don't use shape
-					peak_harmonic[pk] = peaks[pk].freq / (pitch*16);
+					peak_harmonic[pk] = peaks[pk].freq / (wdata.pitch*16);
 				}
 
 				// adjust amplitude to compensate for fewer harmonics at higher pitch
-				amplitude2 = (amplitude * pitch)/(100 << 11);
+				amplitude2 = (wdata.amplitude * wdata.pitch)/(100 << 11);
 
 				if(glottal_flag > 0)
 				{
@@ -1337,22 +1342,22 @@ int Wavegen()
 
 		// mix with sampled wave if required
 		z2 = 0;
-		if(mix_wavefile_ix < n_mix_wavefile)
+		if(wdata.mix_wavefile_ix < wdata.n_mix_wavefile)
 		{
-			if(mix_wave_scale == 0)
+			if(wdata.mix_wave_scale == 0)
 			{
 				// a 16 bit sample
-				c = mix_wavefile[mix_wavefile_ix+1];
-				sample = mix_wavefile[mix_wavefile_ix] + (c * 256);
-				mix_wavefile_ix += 2;
+				c = wdata.mix_wavefile[wdata.mix_wavefile_ix+1];
+				sample = wdata.mix_wavefile[wdata.mix_wavefile_ix] + (c * 256);
+				wdata.mix_wavefile_ix += 2;
 			}
 			else
 			{
 				// a 8 bit sample, scaled
-				sample = (signed char)mix_wavefile[mix_wavefile_ix++] * mix_wave_scale;
+				sample = (signed char)wdata.mix_wavefile[wdata.mix_wavefile_ix++] * wdata.mix_wave_scale;
 			}
-			z2 = (sample * amplitude_v) >> 10;
-			z2 = (z2 * mix_wave_amp)/32;
+			z2 = (sample * wdata.amplitude_v) >> 10;
+			z2 = (z2 * wdata.mix_wave_amp)/32;
 		}
 
 		z1 = z2 + (((total>>8) * amplitude2) >> 13);
@@ -1587,8 +1592,8 @@ static void SetAmplitude(int length, unsigned char *amp_env, int value)
 	else
 		amp_inc = (256 * ENV_LEN * STEPSIZE)/length;
 
-	amplitude = (value * general_amplitude)/16;
-	amplitude_v = (amplitude * wvoice->consonant_ampv * 15)/100;           // for wave mixed with voiced sounds
+	wdata.amplitude = (value * general_amplitude)/16;
+	wdata.amplitude_v = (wdata.amplitude * wvoice->consonant_ampv * 15)/100;           // for wave mixed with voiced sounds
 
 	amplitude_env = amp_env;
 }
@@ -1641,18 +1646,18 @@ if(option_log_frames)
 	}
 }
 #endif
-	if((pitch_env = env)==NULL)
-		pitch_env = env_fall;  // default
+	if((wdata.pitch_env = env)==NULL)
+		wdata.pitch_env = env_fall;  // default
 
-	pitch_ix = 0;
+	wdata.pitch_ix = 0;
 	if(length==0)
-		pitch_inc = 0;
+		wdata.pitch_inc = 0;
 	else
-		pitch_inc = (256 * ENV_LEN * STEPSIZE)/length;
+		wdata.pitch_inc = (256 * ENV_LEN * STEPSIZE)/length;
 
-	SetPitch2(wvoice, pitch1, pitch2, &pitch_base, &pitch_range);
+	SetPitch2(wvoice, pitch1, pitch2, &wdata.pitch_base, &wdata.pitch_range);
 	// set initial pitch
-	pitch = ((pitch_env[0]*pitch_range)>>8) + pitch_base;   // Hz << 12
+	wdata.pitch = ((wdata.pitch_env[0] * wdata.pitch_range) >>8) + wdata.pitch_base;   // Hz << 12
 
 	flutter_amp = wvoice->flutter;
 
@@ -1662,15 +1667,14 @@ if(option_log_frames)
 
 
 
-void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2)
-{//============================================================
+void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *v)
+{//========================================================================
 	int ix;
 	DOUBLEX next;
 	int length2;
 	int length4;
 	int qix;
 	int cmd;
-	voice_t *v;
 	static int glottal_reduce_tab1[4] = {0x30, 0x30, 0x40, 0x50};  // vowel before [?], amp * 1/256
 //	static int glottal_reduce_tab1[4] = {0x30, 0x40, 0x50, 0x60};  // vowel before [?], amp * 1/256
 	static int glottal_reduce_tab2[4] = {0x90, 0xa0, 0xb0, 0xc0};  // vowel after [?], amp * 1/256
@@ -1724,8 +1728,6 @@ if(option_log_frames)
 			break;   // next is not from spectrum, so continue until end of wave cycle
 	}
 
-	v = wvoice;
-
 	// round the length to a multiple of the stepsize
 	length2 = (length + STEPSIZE/2) & ~0x3f;
 	if(length2 == 0)
@@ -1767,7 +1769,7 @@ if(option_log_frames)
 static int Wavegen2(int length, int modulation, int resume, frame_t *fr1, frame_t *fr2)
 {//====================================================================================
 	if(resume==0)
-		SetSynth(length,modulation,fr1,fr2);
+		SetSynth(length, modulation, fr1, fr2, wvoice);
 
 	return(Wavegen());
 }
@@ -1839,35 +1841,46 @@ int WavegenFill(int fill_zeros)
 			{
 				echo_complete -= length;
 			}
-			n_mix_wavefile = 0;
+			wdata.n_mix_wavefile = 0;
+			wdata.prev_was_synth = 0;
 			result = PlaySilence(length,resume);
 			break;
 
 		case WCMD_WAVE:
 			echo_complete = echo_length;
-			n_mix_wavefile = 0;
+			wdata.n_mix_wavefile = 0;
+			wdata.prev_was_synth = 0;
 			result = PlayWave(length,resume,(unsigned char*)q[2], q[3] & 0xff, q[3] >> 8);
 			break;
 
 		case WCMD_WAVE2:
 			// wave file to be played at the same time as synthesis
-			mix_wave_amp = q[3] >> 8;
-			mix_wave_scale = q[3] & 0xff;
-			if(mix_wave_scale == 0)
-				n_mix_wavefile = length*2;
+			wdata.mix_wave_amp = q[3] >> 8;
+			wdata.mix_wave_scale = q[3] & 0xff;
+			if(wdata.mix_wave_scale == 0)
+				wdata.n_mix_wavefile = length*2;
 			else
-				n_mix_wavefile = length;
-			mix_wavefile_ix = 0;
-			mix_wavefile = (unsigned char *)q[2];
+				wdata.n_mix_wavefile = length;
+			wdata.mix_wavefile_ix = 0;
+			wdata.mix_wavefile = (unsigned char *)q[2];
 			break;
 
 		case WCMD_SPECT2:   // as WCMD_SPECT but stop any concurrent wave file
-			n_mix_wavefile = 0;   // ... and drop through to WCMD_SPECT case
+			wdata.n_mix_wavefile = 0;   // ... and drop through to WCMD_SPECT case
 		case WCMD_SPECT:
 			echo_complete = echo_length;
 			result = Wavegen2(length & 0xffff,q[1] >> 16,resume,(frame_t *)q[2],(frame_t *)q[3]);
 			break;
 
+#ifdef INCLUDE_KLATT
+		case WCMD_KLATT2:   // as WCMD_SPECT but stop any concurrent wave file
+			wdata.n_mix_wavefile = 0;   // ... and drop through to WCMD_SPECT case
+		case WCMD_KLATT:
+			echo_complete = echo_length;
+			result = Wavegen_Klatt2(length & 0xffff,q[1] >> 16,resume,(frame_t *)q[2],(frame_t *)q[3]);
+			break;
+#endif
+
 		case WCMD_MARKER:
 			MarkerEvent(q[1],q[2],q[3],out_ptr);
 #ifdef LOG_FRAMES