/
mbrola.c
1056 lines (891 loc) · 30.8 KB
/
mbrola.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
* File: mbrola.c
* Purpose: Diphone-based MBROLA speech synthesizer.
* Authors: Thierry Dutoit and Vincent Pagel
* Email : mbrola@tcts.fpms.ac.be
* Time-stamp: <00/04/11 13:59:18 pagel>
*
* FPMs-TCTS SOFTWARE LIBRARY
*
* Copyright (c) 1995-2018 Faculte Polytechnique de Mons (TCTS lab)
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <https://www.gnu.org/licenses/>.
*
* This file contains an implementation or Belgian patent BE09600524
* and US Patent Nr. 5,987,413
*
* 10/01/97 : Created.
*
* 22/10/97 : incredible bug with maxnconcat
*
* 17/06/98 : Mbrola is now pseudo object code with a This pointer
*
* 31/08/98 : init ola_win to 0 for PCs
* bug for 0ms long phonemes
*
* 04/09/98 : return error code on every floor (to replace catch/throw)
*
* 08/09/98 : correct double memory free in case of reset_Mbrola
* correct leftPhone and rightPhone affectation order for cur_diph
* to avoid memory leak in case of premature ending
*
* 18/05/99 : decrement zero_padding() and add a condition not to overlap
* with phonemes having 0ms length
*
* 10/06/99 : ola_win is zeroed in reset_Mbrola instead of init_Mbrola
* (clean buffer in case of playback interruption)
*
* 05/09/99 : boundcheck on "first" in Concat
*
* 28/03/00 : Rom databases, malloc'ed diphone buffers
* Test in Concat for degenerated case "0ms long phonemes"
*/
#include <math.h>
#include "common.h"
#include "diphone.h"
#include "database.h"
#include "audio.h"
#include "hash_tab.h"
#include "parser.h"
#include "mbrola.h"
void init_Hanning(float* table,int size,float ratio)
/*
* Initialize the Hanning weighting window
* Ratio is used for volume control (Volume is embedded in hanning to
* spare a multiplication later ...)
*/
{
int i;
for (i=0; i<size; i++)
table[i]= (float) (ratio * 0.5 * (1.0 - cos((double)i*2*3.14159265358979323846/(float)size)));
}
void set_voicefreq_Mbrola(Mbrola* mb, uint16 OutFreq)
/* Change the Output Freq and VoiceRatio to change the vocal tract */
{
VoiceFreq(mb)=OutFreq;
VoiceRatio(mb)=(float) VoiceFreq(mb) / (float)Freq(diph_dba(mb));
}
uint16 get_voicefreq_Mbrola(Mbrola* mb)
/* Get output Frequency */
{ return( VoiceFreq(mb)); }
void set_smoothing_Mbrola(Mbrola* mb, bool smoothing)
/* Spectral smoothing or not */
{ smoothing(mb)= smoothing; }
bool get_smoothing_Mbrola(Mbrola* mb)
/* Spectral smoothing or not */
{ return smoothing(mb) ; }
void set_no_error_Mbrola(Mbrola* mb, bool no_error)
/* Tolerance to missing diphones */
{ no_error(mb)=no_error; }
bool get_no_error_Mbrola(Mbrola* mb)
/* Spectral smoothing or not */
{ return no_error(mb); }
void set_volume_ratio_Mbrola(Mbrola* mb, float volume_ratio)
/* Overall volume */
{
init_Hanning( weight(mb), 2*MBRPeriod(diph_dba(mb)), volume_ratio);
volume_ratio(mb)= volume_ratio;
}
float get_volume_ratio_Mbrola(Mbrola* mb)
/* Overall volume */
{ return volume_ratio(mb); }
void set_parser_Mbrola(Mbrola* mb, Parser* parser)
/* drop the current parser for a new one */
{ parser(mb)= parser; }
Mbrola* init_Mbrola(Database* dba)
/*
* Connect the database to the synthesis engine, then initialize internal
* variables. Connect the phonemic command stream later with set_parser_Mbrola
*/
{
Mbrola* mb;
debug_message1("init_Mbrola\n");
mb= (Mbrola*) MBR_malloc(sizeof(Mbrola));
diph_dba(mb) =dba;
/* Allocate buffers */
ola_win(mb) = MBR_malloc( sizeof(float)* MBRPeriod(dba)* 2 );
ola_integer(mb) = MBR_malloc( sizeof(int16)* MBRPeriod(dba)*2 );
weight(mb) = MBR_malloc( sizeof(float)* MBRPeriod(dba)*2 );
/* Default settings ! */
set_voicefreq_Mbrola(mb, Freq(dba)); /* VoiceRatio=1.0 */
set_volume_ratio_Mbrola(mb, 1.0f);
set_smoothing_Mbrola(mb,True);
set_no_error_Mbrola(mb,False);
saturation(mb) =False;
audio_length(mb) =0;
last_time_crumb(mb) =0;
#ifdef LIBRARY
first_call(mb)=True;
#endif
/* prev_diph points to the previous diphone synthesis structure
* and cur_diph points to the current one. While progressing to
* the next diphone, prev_diph memory is resetted the pointers
* are swapped between cur and prev diphones
*/
prev_diph(mb)= init_DiphoneSynthesis(MBRPeriod(dba),
max_frame(dba),
max_samples(dba) );
cur_diph(mb)= init_DiphoneSynthesis(MBRPeriod(dba),
max_frame(dba),
max_samples(dba) );
nb_end(mb)=1000; /* set to high value for first pass in Concat() */
debug_message1("done init_Mbrola\n");
return(mb);
}
void close_Mbrola(Mbrola* mb)
/* Free the memory ! */
{
debug_message1("close_Mbrola\n");
/*
* Right phoneme of prev_diph, and Left phoneme of cur_diph are shared references
* So to avoid double desallocation, scratch one (cur_diph may be scratched in
* case of error)
*/
LeftPhone(cur_diph(mb))=NULL;
close_DiphoneSynthesis(cur_diph(mb));
close_DiphoneSynthesis(prev_diph(mb));
/* Buffers and windows */
MBR_free( ola_win(mb) );
MBR_free( ola_integer(mb) );
MBR_free( weight(mb) );
MBR_free(mb);
debug_message1("done close_Mbrola\n");
}
bool reset_Mbrola(Mbrola* mb)
/*
* Gives initial values to current_diphone (not synthesized anyway)
* -> it will give a first value for prev_diph when we make the first
* NextDiphone call so that cur_diph= _-FirstPhon with lenght1=0
* and prev_diph= _-_ with length2=0
*
* return False in case of error
*/
{
int i;
debug_message1("reset_Mbrola\n");
/* Free residual phonemes
* Right phoneme of prev_diph, and Left phoneme of cur_diph are shared.
* So to avoid double desallocation, scratch one
*/
reset_DiphoneSynthesis(cur_diph(mb));
RightPhone(prev_diph(mb))=NULL;
reset_DiphoneSynthesis(prev_diph(mb));
/* Set dummy cur_diphone with empty value */
LeftPhone(cur_diph(mb)) = init_Phone( sil_phon(diph_dba(mb)), 0.0);
RightPhone(cur_diph(mb)) = init_Phone( sil_phon(diph_dba(mb)), 0.0);
if (!diph_dba(mb)->getdiphone_Database( diph_dba(mb), cur_diph(mb) ))
{
fatal_message( ERROR_DBNOSILENCE,
"_-_ PANIC with %s!\n",
sil_phon(diph_dba(mb)));
return False;
}
/* Indicate we don't generate anything with the first _-_ diphone */
nb_pm(cur_diph(mb))=1;
/* The ola window is null from the start */
for (i=0; i< 2*MBRPeriod(diph_dba(mb)); i++)
ola_win(mb)[i]=0.0f;
#ifdef LIBRARY
/* Indicate that the first call to read_MBR must trigger an initialization */
first_call(mb)=True;
#endif
debug_message1("done reset_Mbrola\n");
return True;
}
StatePhone NextDiphone(Mbrola* mb)
/*
* Reads a phone from the phonetic command buffer and prepares the next
* diphone to synthesize ( prev_diph )
* Return value may be: PHO_EOF,PHO_FLUSH,PHO_OK, PHO_ERROR
*/
{
int len_left, len_right, tot_len;
float wanted_len;
int len_anal; /* Number of sample in last segment analysed */
Phone* my_phone;
StatePhone state;
DiphoneSynthesis *temp;
debug_message1("NextDiphone\n");
if ((state=parser(mb)->nextphone_Parser(parser(mb),&my_phone))!=PHO_OK)
{
debug_message1("done NextDiphone PHO_NOT_OK\n");
return(state);
}
/* length and freq modified if the vocal tract length is not 1.0 */
applyRatio_Phone(my_phone,VoiceRatio(mb));
/* Prev_diph has been used ... release LeftPhone
* don't RightPhone, as reference on it is still active in cur_diph
*/
if (LeftPhone(prev_diph(mb)))
{
close_Phone(LeftPhone(prev_diph(mb))); /* once used, release the phoneme */
LeftPhone(prev_diph(mb))=NULL;
}
/* Shift the diphone being worked on */
/* prev_diph is the one that will be synthesized */
temp=prev_diph(mb);
prev_diph(mb)=cur_diph(mb);
cur_diph(mb)=temp;
/* Initialize minimal info of the new cur_diph */
RightPhone(cur_diph(mb))= my_phone;
LeftPhone(cur_diph(mb)) = RightPhone(prev_diph(mb)); /* Reuse common part ! */
/*
* Load the diphone from the database !!
*/
if ( ! diph_dba(mb)->getdiphone_Database( diph_dba(mb), cur_diph(mb)) )
{
bool success= False; /* then ... we failed */
/* Try to arrange a solution for the missing diphone */
if (no_error(mb))
{
PhonemeName temp_left= name_Phone( LeftPhone( cur_diph(mb)));
PhonemeName temp_right= name_Phone( RightPhone( cur_diph(mb)));
warning_message(ERROR_UNKNOWNSEGMENT,
"Warning: %s-%s unknown, replaced with _-_\n",
name_Phone(LeftPhone( cur_diph(mb))),
name_Phone(RightPhone( cur_diph(mb))));
/* Momentary replacement with _-_ */
name_Phone( LeftPhone( cur_diph(mb)))= sil_phon( diph_dba(mb));
name_Phone( RightPhone( cur_diph(mb)))= sil_phon(diph_dba(mb));
success= diph_dba(mb)->getdiphone_Database( diph_dba(mb), cur_diph(mb));
/* Restore situation */
name_Phone( LeftPhone( cur_diph(mb)))= temp_left;
name_Phone( RightPhone( cur_diph(mb)))= temp_right;
}
if (!success)
{
fatal_message(ERROR_UNKNOWNSEGMENT,
"Fatal error: Unknown recovery for %s-%s segment\n",
name_Phone(LeftPhone(cur_diph(mb))),
name_Phone(RightPhone(cur_diph(mb))));
return PHO_ERROR;
}
}
/*
* Computation of length of 2nd part of prev_diph and of 1st part of current
* one : current phoneme gives current diphone => impose length of 2nd
* part of previous and of 1st part of current one proportionally to their
* relative importance in the database.
* Ex : phoneme sequence = A B C. When synthesizer receives C, B-C is
* scheduled, and the length of B in A-B and B-C are derived. Synthesis of
* A-B can then start.
*/
len_anal = ( nb_frame_diphone(prev_diph(mb)) ) * MBRPeriod(diph_dba(mb));
len_left= len_anal - halfseg_diphone(prev_diph(mb));
len_right= halfseg_diphone(cur_diph(mb)) ;
tot_len=len_left+len_right;
wanted_len= length_Phone( LeftPhone(cur_diph(mb)) );
Length2(prev_diph(mb))= (int) (wanted_len * (float)len_left * (float)Freq(diph_dba(mb))
/ 1000.0f / (float)tot_len) ;
Length1(cur_diph(mb)) = (int) (wanted_len * (float) Freq(diph_dba(mb)) / 1000.0f -
Length2(prev_diph(mb))) ;
debug_message3("done NextDiphone PHO_OK -> %s-%s\n",
name_Phone(RightPhone(prev_diph(mb))),
name_Phone(my_phone));
return(PHO_OK);
}
bool MatchProsody(Mbrola* mb)
/*
* Selects Duplication or Elimination for each analysis OLA frames of
* the diphone we must synthesize (prev_diph). Selected frames must fit
* with wanted pitch pattern and phonemes duration of prev_diph
*
* Return False in case of error
*/
{
int nb_frame; /* Nbr of pitch markers in the database segment*/
float theta1; /* Time warping coefficients */
float theta2=0.0f;
float beta=0.0f;
int halfseg; /* Index (in samples) of center of diphone */
int cur_sample; /* sample offset in synthesis window */
int t; /* time in analysis window */
int limit; /* sample where to stop */
int i;
int k; /* dummy variable !!! */
int len_anal; /* Number of sample in last segment analysed */
int start;
int old_len1; /* Length1 in samples before adjustment */
debug_message1("MatchProsody\n");
/* Modify the length of the 2nd part of a phone if the end of its 1st
part does not correspond to its theoretical value */
start=(int) ( length_Phone(LeftPhone(prev_diph(mb))) -
Length1(prev_diph(mb)) / (Freq(diph_dba(mb))*1000.0f));
old_len1=Length1(prev_diph(mb));
if (old_len1>0)
{
Length1(prev_diph(mb))+= last_time_crumb(mb) ;
/* redraw pitch curve with new length */
for (i=0; i< NPitchPatternPoints(LeftPhone(prev_diph(mb))) ;i++)
if (PitchPattern(LeftPhone(prev_diph(mb)))[i].pos > start)
{
PitchPattern(LeftPhone(prev_diph(mb)))[i].pos= start+
(PitchPattern(LeftPhone(prev_diph(mb)))[i].pos - start) *
Length1(prev_diph(mb))/old_len1;
}
}
nb_frame = nb_frame_diphone(prev_diph(mb));
halfseg = halfseg_diphone(prev_diph(mb));
len_anal = nb_frame * MBRPeriod(diph_dba(mb));
theta1 = ((float) Length1(prev_diph(mb))) / (float) halfseg;
if (len_anal-halfseg!=0)
{
theta2 = ((float) Length2(prev_diph(mb))) / (float) (len_anal-halfseg);
beta = (float) Length1(prev_diph(mb)) - (theta2*halfseg);
}
/*
* Assign to each synthesis frame the corresponding one in segment analysis
*/
frame_number(mb)[0]=1;
frame_pos(mb)[0]=0;
k=1; /* first window pos (sample) */
debug_message3("matchpro %i= %i\n",nb_frame, pmrk_DiphoneSynthesis( prev_diph(mb), 1));
if ( pmrk_DiphoneSynthesis( prev_diph(mb), 1)
& VOICING_MASK )
{
cur_sample = GetPitchPeriod(prev_diph(mb), 0, Freq(diph_dba(mb)));
}
else
{
cur_sample = MBRPeriod(diph_dba(mb));
}
t= MBRPeriod(diph_dba(mb));
/* stride through analysis frames */
for(i=1; i<=nb_frame ;i++)
{
if (t <= halfseg) /* Left phone */
limit = (int) (t*theta1);
else /* Right phone */
limit = (int) (t*theta2 + beta);
/* Stride through synthesis windows */
while (cur_sample <= limit)
{
frame_number(mb)[k]=i; /* NB : first frame number = 1 */
frame_pos(mb)[k]= cur_sample;
k++;
debug_message3("match %i= %i\n", i, pmrk_DiphoneSynthesis( prev_diph(mb), i));
if (pmrk_DiphoneSynthesis( prev_diph(mb), i)
& VOICING_MASK )
{
cur_sample += GetPitchPeriod(prev_diph(mb),cur_sample,Freq(diph_dba(mb)));
}
else
{
cur_sample += MBRPeriod(diph_dba(mb));
}
if (k>=NBRE_PM_MAX)
{
fatal_message( ERROR_PITCHTOOHIGH,
"%s-%s Concat : PANIC, check your pitch :-)\n",
name_Phone(LeftPhone(prev_diph(mb))),
name_Phone(RightPhone(prev_diph(mb))));
return False;
}
}
t=t+MBRPeriod(diph_dba(mb));
}
frame_number(mb)[k]=0; /* End tag */
nb_pm(prev_diph(mb)) = k-1; /* Number of pitchmarks for synthesis */
/* total length that should have been synthesized -last effective sample */
last_time_crumb(mb)+= (old_len1+Length2(prev_diph(mb))) - frame_pos(mb)[k-1] ;
debug_message1("done MatchProsody\n");
return True;
}
void Concat(Mbrola* mb)
/*
* This is a unique feature of MBROLA.
* Smoothes diphones around their concatenation point by making the left
* part fade into the right one and conversely. This is possible because
* MBROLA frames have the same length everywhere.
*
* output : nb_begin, nb_end -> number of stable voiced frames to be used
* for interpolation at the end of Leftphone(prev_diph(mb)) and the beginning
* of RightPhone(prev_diph(mb)).
*/
{
int c;
int begin, end; /* Nbr of voiced frames at begining and end of prev_diph(mb)*/
int first,last; /* Number of the first frame */
int last_frame, first_frame; /* offset in sample for the last and first */
/* frame of concatenation point */
int16 *buff_left; /* speech buffer on left of junction */
int16 *buff_right; /* speech buffer on right of junction */
int i,j;
int cur_sample; /* sample offset in synthesis window */
int maxnconcat;
int limitframe;
debug_message1("Concat\n");
/* We compute the first Olaed frame for cur_diph -> we can't do
* Matchprosody on it yet since we lack pitch points...
* But at least we have a pitch point at 0 for cur_diph
*/
cur_sample = GetPitchPeriod(cur_diph(mb), 0, Freq(diph_dba(mb)));
if (Length1(cur_diph(mb))!=0)
{
float theta= ((float) Length1(cur_diph(mb))) / halfseg_diphone(cur_diph(mb));
/* Use CEIL as first belongs to [1..N] */
first= (int) ceil((double)cur_sample / MBRPeriod(diph_dba(mb)) / theta);
}
else
{
first=0; /* 0 length phonemes */
}
if (first > frame_number(mb)[ nb_pm(cur_diph(mb)) ])
{
first= frame_number(mb)[ nb_pm(cur_diph(mb)) ];
}
last= frame_number(mb)[nb_pm(prev_diph(mb))];
if (
(first!=0) && /* Degenerated case= phoneme with 0ms length */
(pmrk_DiphoneSynthesis( prev_diph(mb), last ) & VOICING_MASK) &&
(pmrk_DiphoneSynthesis( cur_diph(mb), first ) & VOICING_MASK))
{
/* Difference vector beetween LAST and FIRST Voiced frame */
smooth(cur_diph(mb))=True;
/* buffer goes from 0 to NB_SAMPLE while real_frame goes from 1 to N */
last_frame = MBRPeriod(diph_dba(mb)) * ( real_frame(prev_diph(mb))[last] -1 );
buff_left= &buffer(prev_diph(mb))[last_frame];
first_frame = MBRPeriod(diph_dba(mb)) * (real_frame(cur_diph(mb))[first] -1 );
buff_right= &buffer(cur_diph(mb))[first_frame];
/* For the first half, no problem */
for(i=0;i<MBRPeriod(diph_dba(mb));i++)
smoothw(cur_diph(mb))[i] = (int) ((buff_left[i]-buff_right[i]) * weight(mb)[i]);
/* For the second half, reset counters of looped frames */
for(j=0;i<(MBRPeriod(diph_dba(mb))*2); i++,j++)
smoothw(cur_diph(mb))[i] = (int) ((buff_left[j]-buff_right[j]) * weight(mb)[i]);
}
else
smooth(cur_diph(mb))=False;
/*
* Count voiced frames on prev_diph(mb) ( available for smoothing )
* 'begin' for the left part and 'end' for the right part
*/
limitframe= halfseg_diphone(prev_diph(mb)) / MBRPeriod(diph_dba(mb));
maxnconcat=(int) (0.5*nb_pm(prev_diph(mb))+1.0);
if (maxnconcat>6) maxnconcat=6;
begin=1;
while ((pmrk_DiphoneSynthesis(prev_diph(mb), frame_number(mb)[begin]) == V_REG)
&& (frame_number(mb)[begin] <= limitframe)
&& (begin < maxnconcat))
begin++;
/* if VREG VREG ... VREG VTRA : smooth VTRA too */
if ( ( begin>1 )
&& ( pmrk_DiphoneSynthesis(prev_diph(mb),frame_number(mb)[begin]) == V_TRA)
&& ( frame_number(mb)[begin] <= limitframe)
&& ( begin < maxnconcat))
begin++;
nb_begin(mb) = begin -1;
/* Check previous value of nb_end */
if (nb_begin(mb) > nb_end(mb))
nb_begin(mb)=nb_end(mb);
debug_message2("nb_begin %i \n", nb_begin(mb));
end=1;
c=nb_pm(prev_diph(mb))+1;
while ((pmrk_DiphoneSynthesis(prev_diph(mb),frame_number(mb)[c-end]) == V_REG)
&& (frame_number(mb)[c-end] > limitframe)
&& (end < maxnconcat))
end++;
/* if VTRA VREG VREG ... VREG : smooth VTRA too */
if ((end > 1)
&& (pmrk_DiphoneSynthesis(prev_diph(mb),frame_number(mb)[c-end]) == V_TRA)
&& (frame_number(mb)[c-end] > limitframe)
&& (end < maxnconcat))
end++;
nb_end(mb) = end - 1;
/* simplification: just check 1st frame of cur_diph, if not
* voiced, no interpolation at end of prev_diph(mb)
*/
if (pmrk_DiphoneSynthesis(cur_diph(mb), 1) != V_REG)
nb_end(mb)=0;
debug_message2("end %i ", nb_end(mb));
debug_message1("done Concat\n");
}
void FlushFile(Mbrola* mb, int shift, int shift_zero)
/*
* Flush on file what's computed
*/
{
int k;
for (k=0;k<shift;k++)
{
if (ola_win(mb)[k] > 32765)
{
saturation(mb)=True;
ola_integer(mb)[k]=32765;
}
else if (ola_win(mb)[k] < -32765)
{
saturation(mb)=True;
ola_integer(mb)[k]=-32765;
}
else
ola_integer(mb)[k]= (int16) ola_win(mb)[k];
}
/* Amount that has been flushed */
buffer_shift(mb)=shift;
/* Zero padding ! */
zero_padding(mb)= shift_zero;
#ifndef LIBRARY
audio_length(mb)+= write_int16s(ola_integer(mb), shift, output_file);
/* Fill the gap between 2 frames for extra low pitch */
/* No lower limit, but write MBRPeriod buffer each time */
if (shift_zero>0)
{
int shift_mod; /* Modulo for shift zero */
int written;
if (shift_zero>2*MBRPeriod(diph_dba(mb)))
shift_mod=2*MBRPeriod(diph_dba(mb));
else
shift_mod=shift_zero;
for (k=0; k<shift_mod; k++) ola_integer(mb)[k]=0;
while (shift_zero>2*MBRPeriod(diph_dba(mb)))
{
written= write_int16s(ola_integer(mb), 2*MBRPeriod(diph_dba(mb)), output_file);
buffer_shift(mb)+=written;
audio_length(mb)+=written;
shift_zero-=written;
}
written= write_int16s(ola_integer(mb),shift_zero, output_file);
audio_length(mb)+=written;
buffer_shift(mb)+=written;
}
#endif
}
void OverLapAdd(Mbrola* mb, int frame)
/*
* OLA routine
*/
{
int k;
float correction; /* Energy correction factor */
int end_window, add_window;
int lim_smooth; /* Beyond this limit -> left smoothing */
float tmp;
FrameType type; /* Frame type */
int shift_zero; /* Noman's land between 2 ola filled with 0 */
int shift; /* Shift between pulses */
debug_message1("OverLapAdd\n");
shift = frame_pos(mb)[frame]-frame_pos(mb)[frame-1];
if ((correction = (float)shift/(float)MBRPeriod(diph_dba(mb)))>=1) correction=1.0f;
/* Keep nothing of previous frames as there's no overlap */
shift_zero= shift - 2*MBRPeriod(diph_dba(mb));
if (shift_zero>0)
shift= 2*MBRPeriod(diph_dba(mb));
end_window = 2*MBRPeriod(diph_dba(mb)) - shift;
add_window =MBRPeriod(diph_dba(mb)) * (real_frame(prev_diph(mb))[frame_number(mb)[frame]]-1);
lim_smooth = nb_pm(prev_diph(mb))-nb_end(mb);
/* Flush on file what's flushable */
FlushFile(mb,shift,shift_zero);
if (saturation(mb))
{
warning_message(WARNING_SATURATION,
"Saturation on %s-%s\n",
name_Phone(LeftPhone(prev_diph(mb))),
name_Phone(RightPhone(prev_diph(mb))));
saturation(mb)=False;
}
/* !! SHIFTING CAN BE REMOVED IN CASE OF STATIC OUTPUT BUFFER !! */
/* shift the ola window and completion with 0 */
memmove(&ola_win(mb)[0], &ola_win(mb)[shift], end_window*sizeof(ola_win(mb)[0]));
for(k=end_window; k<2*MBRPeriod(diph_dba(mb)) ; k++)
ola_win(mb)[k]=0.0f;
/* Two kinds of OLA depending if the frame is unvoiced */
type= pmrk_DiphoneSynthesis(prev_diph(mb), frame_number(mb)[frame] );
debug_message5("Frame type=%i Number=%i Real=%i Smooth=%i\n",
type,
frame_number(mb)[frame],
real_frame(prev_diph(mb))[frame_number(mb)[frame]],
smooth(prev_diph(mb)));
if (! (type & VOICING_MASK))
{ /* UNVOICED FRAME => PLAY BACK */
if ( (frame<=nb_begin(mb))||
(frame>lim_smooth))
warning_message(ERROR_SMOOTHING,"Smooth Panic! Stay cool");
/* Check if we're duplicating an unvoiced frame */
if ( odd(mb) &&
(frame_number(mb)[frame] == frame_number(mb)[frame-1]))
{
/* reverse every second duplicated UV frame */
add_window=add_window+2*MBRPeriod(diph_dba(mb))-1;
for (k=0; k< 2*MBRPeriod(diph_dba(mb)) ; k++)
{
tmp = weight(mb)[k] * (float) buffer(prev_diph(mb))[add_window - k];
/* Energy correction */
tmp *= correction;
ola_win(mb)[k] += tmp;
}
}
else /* Don't reverse the unvoiced frame */
{
for (k=0; k< 2*MBRPeriod(diph_dba(mb)) ; k++)
{
tmp = weight(mb)[k] * (float) buffer(prev_diph(mb))[add_window + k];
/* Energy correction */
tmp *= correction;
ola_win(mb)[k] += tmp;
}
}
}
else
{
/* Meeep meeep, you're entering a restrictead area, helmet is
* recommended if you don't have rights to exploit Belgian
* patent BE09600524 or US Patent Nr. 5,987,413
*
* Voiced frame -> autoloop ! MBROLA unique feature !!
* Many case depending on smoothing or not
*/
if ((frame<=nb_begin(mb)) &&
smooth(prev_diph(mb)) &&
smoothing(mb) )
/* Left smoothing */
{
float smooth_left = (float)(nb_begin(mb)-frame+1) / (2*(float)nb_begin(mb));
for (k=0; k<MBRPeriod(diph_dba(mb)) ; k++)
ola_win(mb)[k] += correction *
( weight(mb)[k] * (float)buffer(prev_diph(mb))[add_window + k]
+ smooth_left * smoothw(prev_diph(mb))[k] );
for ( ; k<2*MBRPeriod(diph_dba(mb)) ; k++)
ola_win(mb)[k] += correction *
( weight(mb)[k] * (float) buffer(prev_diph(mb)) [add_window + k - MBRPeriod(diph_dba(mb))]
+ smooth_left * smoothw(prev_diph(mb))[k] );
}
else if ( (frame>lim_smooth) &&
smooth(cur_diph(mb)) &&
smoothing(mb) )
/* Right smoothing */
{
float smooth_right= (float)(nb_end(mb)-(nb_pm(prev_diph(mb))-frame))
/(2*(float)nb_end(mb));
for (k=0; k<MBRPeriod(diph_dba(mb)) ; k++)
ola_win(mb)[k] += correction *
( weight(mb)[k] * (float)buffer(prev_diph(mb))[add_window + k]
- smooth_right * smoothw(cur_diph(mb)) [k]);
for ( ; k<2*MBRPeriod(diph_dba(mb)) ; k++)
ola_win(mb)[k] += correction *
( weight(mb)[k] * (float) buffer(prev_diph(mb)) [add_window + k - MBRPeriod(diph_dba(mb))]
- smooth_right * smoothw(cur_diph(mb))[k]);
}
else
/* No smoothing */
{
for (k=0; k<MBRPeriod(diph_dba(mb)) ; k++)
ola_win(mb)[k] += correction * weight(mb)[k] *
(float)buffer(prev_diph(mb))[add_window + k];
for ( ; k<2*MBRPeriod(diph_dba(mb)) ; k++)
ola_win(mb)[k] += correction * weight(mb)[k] *
(float) buffer(prev_diph(mb))[add_window + k - MBRPeriod(diph_dba(mb))];
}
}
odd(mb)= !odd(mb); /* Flip flop for NV frames */
debug_message1("done OverLapAdd\n");
}
#ifdef LIBRARY
/* LIBRARY mode: synthesis driven by the output */
int readtype_Mbrola(Mbrola* mb, void *buffer_out, int nb_wanted, AudioType sample_type)
/*
* Reads nb_wanted samples in an audio buffer
* Returns the effective number of samples read
*
* negative means error code
*/
{
int to_go= nb_wanted; /* number of samples to go */
int nb_move;
if (first_call(mb))
{
odd(mb)=0;
eaten(mb)=0;
buffer_shift(mb)=0;
zero_padding(mb)=0;
frame_counter(mb)=1;
if (!reset_Mbrola(mb))
return lasterr_code;
/* Test if there is something in the buffer */
switch ( NextDiphone(mb) )
{
case PHO_OK: /* go ahead */
break;
case PHO_ERROR: /* return error code */
return lasterr_code;
case PHO_EOF:
case PHO_FLUSH: /* Flush or EOF, 0 samples */
return 0;
default:
/* panic ! */
fatal_message(ERROR_NEXTDIPHONE,"NextDiphone PANIC %i!\n");
return ERROR_NEXTDIPHONE ;
}
first_call(mb)=False;
}
while (to_go>0)
{
/* Zero padding between 2 Ola frames for extra low pitch */
if (zero_padding(mb)>0)
{
int nb_generated;
if ( zero_padding(mb) > to_go)
nb_generated= to_go;
else
nb_generated= zero_padding(mb);
/* decrement in case the requested buffer is very small */
zero_padding(mb)-= nb_generated;
buffer_out= zero_convert( buffer_out,nb_generated, sample_type );
if (!buffer_out)
return lasterr_code;
to_go -= nb_generated;
}
/* Available frames from previous OLA */
if (to_go > (buffer_shift(mb)-eaten(mb)))
nb_move= buffer_shift(mb)-eaten(mb);
else
nb_move= to_go;
buffer_out= move_convert( buffer_out, &ola_integer(mb)[eaten(mb)], nb_move, sample_type );
if (!buffer_out)
return lasterr_code;
to_go-= nb_move;
eaten(mb)+= nb_move;
if (to_go<=0)
break;
/* Still some frames available in the same diphone ? */
if (frame_counter(mb)<nb_pm(prev_diph(mb)))
frame_counter(mb)++;
else
{ /* Else, nextdiphone */
StatePhone stream_state;
stream_state= NextDiphone(mb);
if (stream_state != PHO_OK)
{
/* handle errors in the parser */
if (stream_state == PHO_ERROR)
return lasterr_code;
/* Flush or EOF */
if (stream_state == PHO_FLUSH)
first_call(mb)=True;
break;
}
if ( !MatchProsody(mb) )
return lasterr_code;
Concat(mb);
odd(mb)=0;
frame_counter(mb)=1;
}
eaten(mb)=0;
/* condition against phonemes with 0 length */
if (frame_counter(mb)<=nb_pm(prev_diph(mb)))
OverLapAdd(mb, frame_counter(mb));
}
/* old C++ catch throw "} catch(int ret) { return ret; }" */
return(nb_wanted - to_go);
}
#else
/*
* STANDALONE MODE: Synthesis driven by the input
*
* WARNING: we don't care of error value returned, as fatal_messages
* are really fatal in standalone, call exit()
*/
void oneshot_Mbrola(Mbrola* mb)
/*
* Completely perform MBROLA synthesis of prev_diph(mb)
* Different from audio chunks synthesized on demand in library mode
*/
{
debug_message1("Mbrola\n");
MatchProsody(mb); /* we're in standalone */
debug_message3("%s-%s Concat : ",
left_diphone(mb, prev_diph(mb)),
right_diphone(mb, prev_diph(mb)));
Concat(mb);
#ifdef DEBUG
{
int frame_counter;
debug_message4("%s-%s NbPM :%u ",
left_diphone(mb,prev_diph(mb)),
right_diphone(mb,prev_diph(mb)),
nb_pm(prev_diph(mb)));
for (frame_counter=1; frame_counter<=nb_pm(prev_diph(mb)); frame_counter++)
debug_message4("(%u %u %u) ",
frame_number(mb)[frame_counter],
pmrk_DiphoneSynthesis(prev_diph(mb),
real_frame(prev_diph(mb))[frame_number(mb)[frame_counter]]),
frame_pos(mb)[frame_counter]-frame_pos(mb)[frame_counter-1]);
debug_message1("\nSamples : ");
}
#endif
odd(mb)=False;
for (frame_counter(mb)=1; frame_counter(mb)<=nb_pm(prev_diph(mb)); frame_counter(mb)++)
{
#ifdef SIGNAL
if (must_flush) /* test if premature return requested (reset signal) */
return;
#endif
OverLapAdd(mb,frame_counter(mb));
}
#ifdef DEBUG
debug_message1("\n");
debug_message4("DEBUG: LeftPhone=%s RightPhone=%s genere %i\n",
left_diphone(mb , prev_diph(mb)),
right_diphone(mb , prev_diph(mb)),
frame_counter(mb));
debug_message3("DEBUG: LeftPhone=%i RightPhone=%i\n",
Length1(prev_diph(mb)),
Length2(prev_diph(mb)));
debug_message3("DEBUG: Patternpointsleft=%i Right=%i \n",