#include <phrase.h>
Collaboration diagram for phrase:
Public Methods | |
| uint | MeanF0 () |
| returns the mean val of all f0Vals. More... | |
| void | calcMeanF0 () |
| calculates the mean val of all f0Vals. More... | |
| double | F0Range () |
| returns range of f0Vals in SemiTones. More... | |
| void | printSyllableVec (ostream &outFile) |
| prints names of phones to outFile (debugging). More... | |
| void | readPhoFile (istream &inFile, globalsT &kg) |
| reads in phones for syllables from inFile. More... | |
| void | readPhoList (istream &inFile, globalsT &kg) |
| assignes only names of phones and syllable borders. More... | |
| void | readF0FromDB (string diphonBase, globalsT &glob) |
| assignes only f0-vals. More... | |
| void | readPhonData (string diphonBase, globalsT &glob) |
| reads in amplitudes, formantfreqs and formantbandwidths from diphon-database. More... | |
| void | modelKlattParameters (globalsT &glob) |
| model voice-source and articulation-tract filter parameters for each phone for normal voice. More... | |
| void | modelFeatures (globalsT &kg) |
| model voice-source and articulation-tract filter parameters for each phone according to it's assigned features. More... | |
| void | printPhonMeanF0 (ostream &outFile) |
| prints means of f0-contour per phon to outFile (debugging). More... | |
| void | printPhonFeatures (ostream &outFile) |
| prints phonfeatures to outFile (debugging). More... | |
| void | printPhoFile (ostream &outFile, globalsT glob) |
| prints pho-file (mbrola-format). More... | |
| void | printLabelFile (ostream &outFile, globalsT globals) |
| prints label-file (esps-format). More... | |
| void | printKlattFile (ostream &outFile) |
| prints klatt-file (iles&simmons-format). More... | |
| void | printSenSynFile (globalsT globals, ostream &outFile) |
| prints klatt-file (sensymetrics-format). More... | |
| void | makeTest (globalsT globals, uint dur) |
| fill a phrase with one syllable with one phone of duration dur (in msec) with parameters taken from globals. More... | |
| void | changeSpeechRate (uint rate, globalsT globals, int method) |
| changes duration of all phones. More... | |
| void | changeDurStressed (uint rate, uint both, globalsT globals) |
| change duration of stressed syllables. More... | |
| void | changeDurNonStressed (uint rate, uint both, globalsT globals) |
| change duration of unstressed syllables. More... | |
| void | changeDurManner (PHON_MANNER manner, uint rate, globalsT globals) |
| change duration of specified phone-manners. More... | |
| void | changeStressedInt (int dBVal, uint both, globalsT globals) |
| change intensity of stressed syllables. More... | |
| void | changeMeanF0 (uint rate, globalsT glob) |
| all nonzero f0Vals get multiplied by <rate> in percent. More... | |
| void | smootheF0 (uint winsize, globalsT glob) |
| smoothing function: all f0Vals get centerFiltered around winsize. More... | |
| void | changeStressedAvrF0 (uint rate, uint both, globalsT glob) |
| change mean pitch of stressed syllables. More... | |
| void | changePreStressedAvrF0 (uint rate, uint both, globalsT glob) |
| change avr pitch of the syllables before stressed syllables. More... | |
| void | changeF0Range (uint rate, globalsT glob) |
| changes the F0-range of phrase. More... | |
| void | changeVariability (uint rate, globalsT glob) |
| changes the F0-variability. More... | |
| void | changePitchContourStressed (CONTOUR_TYPE contour, uint gradient, uint both, globalsT glob) |
| changes pitch-contour of stressed syllables to <contour>. More... | |
| void | changeLastSylContour (CONTOUR_TYPE contour, uint gradient, globalsT glob) |
| changes pitch-contour of last syllable to <contour>. More... | |
| void | changePhraseContour (CONTOUR_TYPE contour, uint rate, globalsT glob) |
| changes pitch-contour of whole phrase to <contour>. More... | |
| void | changePhraseContourToAccent (uint rate, globalsT glob) |
| changes pitch-contour of regions of phrase divided by phraseaccents to <contour>. More... | |
| void | addWave (uint accRate, uint nonAccRate, uint method, globalsT glob) |
| The intonation-contour gets changed according to the wave-model. More... | |
| void | overshootVowels (ACCENT_TYPE accent, uint rate, globalsT glob) |
| enhance the vowel precision. More... | |
| void | undershootVowels (ACCENT_TYPE accent, uint rate, globalsT glob) |
| blur the vowel precision. More... | |
| void | spread (uint rate, globalsT glob) |
| adds spread Lips characteristices. More... | |
| void | rounded (uint rate, globalsT glob) |
| adds rounded Lips characteristices. More... | |
| void | headSize (uint rate, globalsT glob) |
| experiment: shift formant-values to model different head-sizes. More... | |
| void | coarticulation (uint rate, CO_TYPE coType, globalsT glob) |
| model coarticulation effects. More... | |
| void | assimilateConsonants (globalsT glob) |
| should model assimilation of voicing. More... | |
| void | addJitter (ACCENT_TYPE accent, uint rate, uint method, globalsT glob) |
| add Jitter (F0-Flutter). More... | |
| void | addBreathy (uint rate, globalsT glob) |
| simulates breathy phonation at all vowels. More... | |
| void | addTense (uint rate, globalsT glob) |
| simulates tense phonation at all vowels. More... | |
| void | addWhisper (uint rate, globalsT glob) |
| simulates whispery phonation at all vowels. More... | |
| void | addLaryngealized (uint rate, globalsT glob) |
| simulates laryngealization (creaky voice) at first half of vowels after unvoiced sounds. More... | |
| void | addCreaky (uint rate, globalsT glob) |
| simulates creaky phonation at all vowels. More... | |
| void | addFalsett (uint rate, globalsT glob) |
| simulates falsetto phonation at all vowels. More... | |
Private Methods | |
| void | calcDur () |
Private Attributes | |
| vector<syllable> | syllableVec |
| uint | meanF0 |
| uint | dur |
It consists of a list of syllable . All changes made are initiated by phrase. The methods can be categorized into
|
|
returns range of f0Vals in SemiTones.
00048 {
00049 uint minf0=A_HIGH_INT, maxf0=0;
00050 for (uint syl=0; syl<syllableVec.size(); syl++) {
00051 for (uint pho=0; pho<syllableVec[syl].PhonVec().size(); pho++) {
00052 if (syllableVec[syl].PhonVec()[pho].Voiced()) {
00053 for (uint f0v=0; f0v<syllableVec[syl].PhonVec()[pho].F0Vec().size(); f0v++) {
00054 if (minf0 > syllableVec[syl].PhonVec()[pho].F0Vec()[f0v] &&
00055 syllableVec[syl].PhonVec()[pho].F0Vec()[f0v]>0)
00056 minf0 = syllableVec[syl].PhonVec()[pho].F0Vec()[f0v];
00057 if (maxf0 < syllableVec[syl].PhonVec()[pho].F0Vec()[f0v])
00058 maxf0 = syllableVec[syl].PhonVec()[pho].F0Vec()[f0v];
00059 }
00060 }
00061 }
00062 }
00063 return Hz2ST(minf0, maxf0);
00064 }
|
|
|
returns the mean val of all f0Vals.
00026 {
00027 calcMeanF0();
00028 return meanF0;
00029 }
|
|
|
simulates breathy phonation at all vowels.
01507 {
01508 phon *pP;
01509 if (rate > 100)
01510 error("phrase::addBreathy: rate>100\n",-5);
01511 for (uint syl=0; syl<syllableVec.size(); syl++) {
01512 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01513 pP = &syllableVec[syl].PhonVec()[pho];
01514 // if ((*pP).Voiced() && (*pP).Manner()!=stop_voiced && (*pP).Name()!="h") {
01515 if ((*pP).Type() == vowel || (*pP).Manner() == nasal) {
01516 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01517 tmpPhonFeatures.breathy = true;
01518 tmpPhonFeatures.brtStart = 0;
01519 tmpPhonFeatures.brtStop = (*pP).FrameNum();
01520 tmpPhonFeatures.brtRate = rate;
01521 (*pP).setPhonFeatures(tmpPhonFeatures);
01522 }
01523 }
01524 }
01525 if(glob.verbose)
01526 cerr << "added breathy voice with rate: "<<rate<<"\n";
01527 }
|
|
|
simulates creaky phonation at all vowels.
01619 {
01620 phon *pP;
01621 if (rate > 100)
01622 error("phrase::addCreaky: rate>100\n",-5);
01623 for (uint syl=0; syl<syllableVec.size(); syl++) {
01624 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01625 pP = &syllableVec[syl].PhonVec()[pho];
01626 if ((*pP).Voiced()) {
01627 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01628 tmpPhonFeatures.creaky = true;
01629 tmpPhonFeatures.creakyStart = 0;
01630 tmpPhonFeatures.creakyStop = (*pP).FrameNum();
01631 tmpPhonFeatures.creakyRate = rate;
01632 (*pP).setPhonFeatures(tmpPhonFeatures);
01633 }
01634 }
01635 }
01636 calcMeanF0();
01637 if (glob.verbose)
01638 cerr << "added creaky voice at rate "<<rate<<" \n";
01639 }
|
|
|
simulates falsetto phonation at all vowels.
01648 {
01649 phon *pP;
01650 if (rate > 100)
01651 error("phrase::addFalsett: rate>100\n",-5);
01652 for (uint syl=0; syl<syllableVec.size(); syl++) {
01653 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01654 pP = &syllableVec[syl].PhonVec()[pho];
01655 if ((*pP).Voiced()) {
01656 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01657 tmpPhonFeatures.falsett = true;
01658 tmpPhonFeatures.falsettRate = rate;
01659 (*pP).setPhonFeatures(tmpPhonFeatures);
01660 }
01661 }
01662 }
01663 calcMeanF0();
01664 if (glob.verbose)
01665 cerr << "added falsett voice at rate "<<rate<<" \n";
01666 }
|
|
|
add Jitter (F0-Flutter). Simulates Jitter at all voiced phones in syllables with stress-type=accent. If method = 1, all nonzero f0Vals get displaced about rate alternating down and up. (0 > rate < 10) If method = 2, the klattSyn parameter skew/FL is set to rate (0 > rate < 300)
01481 {
01482 phon *pP;
01483 for (uint syl=0; syl<syllableVec.size(); syl++) {
01484 if (syllableVec[syl].Accent()==accent) {
01485 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01486 pP = &syllableVec[syl].PhonVec()[pho];
01487 if ((*pP).Type() == vowel) {
01488 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01489 tmpPhonFeatures.jitter = true;
01490 tmpPhonFeatures.jitterRate = rate;
01491 tmpPhonFeatures.jitterMethod = method;
01492 (*pP).setPhonFeatures(tmpPhonFeatures);
01493 }
01494 }
01495 }
01496 }
01497 if(glob.verbose)
01498 cerr << "added jitter with rate: "<<rate<<"\n";
01499 }
|
|
|
simulates laryngealization (creaky voice) at first half of vowels after unvoiced sounds.
01586 {
01587 phon *pP;
01588 bool ok;
01589 if (rate > 100)
01590 error("phrase::addLaryngealized: rate>100\n",-5);
01591 for (uint syl=0; syl<syllableVec.size(); syl++) {
01592 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01593 pP = &syllableVec[syl].PhonVec()[pho];
01594 if ((*pP).Type()==vowel && ok) {
01595 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01596 tmpPhonFeatures.creaky = true;
01597 tmpPhonFeatures.creakyStart = 0;
01598 tmpPhonFeatures.creakyStop = (*pP).FrameNum()/2;
01599 tmpPhonFeatures.creakyRate = rate;
01600 (*pP).setPhonFeatures(tmpPhonFeatures);
01601 }
01602 if (!(*pP).Voiced())
01603 ok = true;
01604 else
01605 ok = false;
01606 }
01607 }
01608 if (glob.verbose)
01609 cerr << "added laryngealized voice at rate "<<rate<<" \n";
01610 calcMeanF0();
01611 }
|
|
|
simulates tense phonation at all vowels.
01535 {
01536 phon *pP;
01537 if (rate > 100)
01538 error("phrase::addWhisper: rate>100\n",-5);
01539 for (uint syl=0; syl<syllableVec.size(); syl++) {
01540 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01541 pP = &syllableVec[syl].PhonVec()[pho];
01542 if ((*pP).Type()==vowel) {
01543 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01544 tmpPhonFeatures.tense = true;
01545 tmpPhonFeatures.tenseRate = rate;
01546 (*pP).setPhonFeatures(tmpPhonFeatures);
01547 }
01548 }
01549 }
01550 if (glob.verbose)
01551 cerr << "added tensed voice at rate "<<rate<<" \n";
01552 }
|
|
|
The intonation-contour gets changed according to the wave-model. The whole intonationContour is changed: the phrase-accented syllables get displaced according to accRate, the syllable between phrase-accents get displaced according to nonAccRate and the other syllables get displaced by interpolation. The method decides about the interpolation,
01181 {
01182 vector<uint> tmpF0Vec, indexList, targetList;
01183 phon *pP;
01184 uint frameCounter=0;
01185
01186
01187 if (method==LINEAR) {
01188 indexList.push_back(0);
01189 targetList.push_back(syllableVec[0].MeanF0());
01190 // read in target vals and indices
01191 for (uint syl=0; syl<syllableVec.size(); syl++) {
01192 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01193 // cout << (*pP).Name()<<" "<<indexList.size()<<"\n";
01194 pP = &syllableVec[syl].PhonVec()[pho];
01195 if (syllableVec[syl].Accent()==sentence && (*pP).Type()==vowel) {
01196 if (indexList.size()>1) {
01197 uint actIndex = frameCounter+(*pP).FrameNum()/2;
01198 uint lastIndex = indexList[indexList.size()-1];
01199 indexList.push_back(lastIndex+((actIndex-lastIndex)/2));
01200 targetList.push_back(meanF0*nonAccRate/100);
01201 }
01202 indexList.push_back(frameCounter+(*pP).FrameNum()/2);
01203 targetList.push_back(syllableVec[syl].MeanF0()*accRate/100);
01204 }
01205 for (uint i=0; i<(*pP).FrameNum(); i++)
01206 tmpF0Vec.push_back((*pP).F0Vec()[i]);
01207 frameCounter += (*pP).FrameNum();
01208 }
01209 }
01210 indexList.push_back(frameCounter);
01211 targetList.push_back(syllableVec[syllableVec.size()-1].MeanF0());
01212
01213 // for (uint i=0; i<indexList.size(); i++)
01214 // cout <<indexList[i]<<" "<< targetList[i] <<"\n";
01215
01216 // modify F0Vals
01217 for (uint i=0; i<indexList.size()-1; i++) {
01218 uint startIndex=indexList[i];
01219 uint endIndex=indexList[i+1];
01220 uint startVal=targetList[i];
01221 uint endVal=targetList[i+1];
01222
01223 // count voiced Frames
01224 uint vocCounter = 0;
01225 for (uint k=startIndex; k<endIndex; k++)
01226 if (tmpF0Vec[k] != 0)
01227 vocCounter++;
01228 if (startVal<endVal) {
01229 double step = (int) (endVal-startVal) / (double) vocCounter;
01230 // cout <<vocCounter<<" "<<step<<" "<<startVal<<" "<<endVal<<"\n";
01231 vocCounter = 0;
01232 for (uint k=startIndex; k<endIndex; k++) {
01233 if (tmpF0Vec[k] != 0) {
01234 tmpF0Vec[k] = startVal+(uint) (vocCounter*step);
01235 vocCounter++;
01236 }
01237 }
01238 } else if (startVal > endVal) {
01239 double step = (int) (startVal-endVal) / (double) vocCounter;
01240 // cout <<vocCounter<<" "<<step<<" "<<startVal<<" "<<endVal<<"\n";
01241 vocCounter = 0;
01242 for (uint k=startIndex; k<endIndex; k++) {
01243 if (tmpF0Vec[k] != 0) {
01244 tmpF0Vec[k] = startVal-(uint) (vocCounter*step);
01245 vocCounter++;
01246 }
01247 }
01248 }
01249
01250 }
01251 // assign new f0 vals to phones
01252 frameCounter = 0;
01253 for (uint syl=0; syl<syllableVec.size(); syl++) {
01254 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01255 pP = &syllableVec[syl].PhonVec()[pho];
01256 vector<uint> tmpValVec(0);
01257 for (uint i=0; i<(*pP).FrameNum(); i++) {
01258 tmpValVec.push_back(tmpF0Vec[frameCounter]);
01259 frameCounter++;
01260 }
01261 (*pP).setF0Vec(tmpValVec, glob);
01262 }
01263 }
01264 //cout <<indexList.size()<<"\n";
01265 } else if (method==SYLLABLE_BASED) {
01266 indexList.push_back(0);
01267 targetList.push_back(100);
01268
01269 for (uint syl=0; syl<syllableVec.size(); syl++) {
01270 if (syllableVec[syl].Accent()==sentence) {
01271 if (indexList.size()>1) {
01272 uint actIndex = syl;
01273 uint lastIndex = indexList[indexList.size()-1];
01274 indexList.push_back(lastIndex+((actIndex-lastIndex)/2));
01275 targetList.push_back(nonAccRate);
01276 }
01277 indexList.push_back(syl);
01278 targetList.push_back(accRate);
01279 }
01280 }
01281 indexList.push_back(syllableVec.size());
01282 targetList.push_back(100);
01283
01284 for (uint i=0; i<indexList.size()-1; i++) {
01285 uint startIndex=indexList[i];
01286 uint endIndex=indexList[i+1];
01287 uint startVal=targetList[i];
01288 uint endVal=targetList[i+1];
01289
01290 // cout << startIndex<<" "<<startVal<<" "<<endIndex<<" "<<endVal<<"\n";
01291
01292 if (startVal<endVal) {
01293 double step = (double) (endVal-startVal) / (double) (endIndex-startIndex);
01294 for (uint syl=startIndex; syl<endIndex; syl++) {
01295 startVal += (uint) step;
01296 // cout << startVal<<" "<<step<<"\n";
01297 syllableVec[syl].changeMeanF0(startVal);
01298 }
01299 } else if (startVal>endVal) {
01300 double step = (double) (startVal-endVal) / (double) (endIndex-startIndex);
01301 for (uint syl=startIndex; syl<endIndex; syl++) {
01302 startVal -= (uint) step;
01303 // cout << startVal<<" "<<step<<"\n";
01304 syllableVec[syl].changeMeanF0(startVal);
01305 }
01306 }
01307 }
01308
01309 } else
01310 error("phrase::addWave: method not implemented\n", -3);
01311 if (glob.verbose)
01312 cerr << "changed f0-countour to wave model at accRate "<<accRate<<" and nonAccRate "<<nonAccRate<<"\n";
01313 calcMeanF0();
01314 }
|
|
|
simulates whispery phonation at all vowels.
01560 {
01561 phon *pP;
01562 if (rate > 100)
01563 error("phrase::addWhisper: rate>100\n",-5);
01564 for (uint syl=0; syl<syllableVec.size(); syl++) {
01565 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01566 pP = &syllableVec[syl].PhonVec()[pho];
01567 if ((*pP).Voiced()) {
01568 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01569 tmpPhonFeatures.whisper = true;
01570 tmpPhonFeatures.whisperRate = rate;
01571 (*pP).setPhonFeatures(tmpPhonFeatures);
01572 }
01573 }
01574 }
01575 if (glob.verbose)
01576 cerr << "added whispery voice at rate "<<rate<<" \n";
01577 }
|
|
|
should model assimilation of voicing. Adds voicing to unvoiced fricatives in VCV-environment. Doesn't work very well.
01443 {
01444 PHON_TYPE prevT, nextT;
01445 phon *pP;
01446 for (uint syl=0; syl<syllableVec.size(); syl++) {
01447 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01448 pP = &syllableVec[syl].PhonVec()[pho];
01449 if (pho==syllableVec[syl].PhonNum()-1)
01450 nextT = syllableVec[syl+1].PhonVec()[0].Type();
01451 else
01452 nextT = syllableVec[syl].PhonVec()[pho+1].Type();
01453
01454 if (!(syl==0&&pho==0) && !(syl==syllableVec.size()-1 &&
01455 pho==syllableVec[syl].PhonNum()-1))
01456 if ((*pP).Manner() == fricative_voiceless && prevT==vowel && nextT==vowel) {
01457 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01458 tmpPhonFeatures.voiced = true;
01459 tmpPhonFeatures.vocStart = 0;
01460 tmpPhonFeatures.vocStop = (*pP).FrameNum();
01461 (*pP).setPhonFeatures(tmpPhonFeatures);
01462 }
01463 prevT = (*pP).Type();
01464 }
01465 }
01466 if(glob.verbose)
01467 cerr << "assimilated consonants in VCV-environment\n";
01468 }
|
|
|
00068 {
00069 uint durSum=0;
00070 for (uint syl=0; syl<syllableVec.size(); syl++) {
00071 for (uint pho=0; pho<syllableVec[syl].PhonVec().size(); pho++) {
00072 durSum += syllableVec[syl].PhonVec()[pho].Dur();
00073 }
00074 }
00075 dur = durSum;
00076 }
|
|
|
calculates the mean val of all f0Vals. To be done after an intonation change or falsetto or creaky voice (they change pitch also) 00031 {
00032 uint f0Sum=0, f0Num=0;
00033 for (uint syl=0; syl<syllableVec.size(); syl++) {
00034 syllableVec[syl].calcMeanF0();
00035 for (uint pho=0; pho<syllableVec[syl].PhonVec().size(); pho++) {
00036 if (syllableVec[syl].PhonVec()[pho].Voiced()) {
00037 for (uint f0v=0; f0v<syllableVec[syl].PhonVec()[pho].F0Vec().size(); f0v++) {
00038 f0Sum += syllableVec[syl].PhonVec()[pho].F0Vec()[f0v];
00039 }
00040 f0Num += syllableVec[syl].PhonVec()[pho].FrameNum();
00041 }
00042 }
00043 }
00044
00045 meanF0 = f0Num>0 ? (uint) f0Sum/f0Num : 0;
00046 }
|
|
|
change duration of specified phone-manners.
00701 {
00702 phon *pP;
00703 if (rate < MIN_SPEECHRATE || rate > MAX_SPEECHRATE)
00704 error ("phrase::changeDurManner: rate < MIN_SPEECHRATE || rate > MAX_SPEECHRATE", -3);
00705 for (uint syl=0; syl<syllableVec.size(); syl++) {
00706 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00707 if (!(syl==0&&pho==0) &&
00708 !(syl==syllableVec.size()-1&&pho==syllableVec[syl].PhonNum()-1)) {
00709 pP = &syllableVec[syl].PhonVec()[pho];
00710 if ((*pP).Manner()==manner) {
00711 uint newDur = (*pP).Dur() * rate/100;
00712 (*pP).changeDur(newDur, glob);
00713 }
00714 }
00715 }
00716 }
00717 calcDur();
00718 if (glob.verbose)
00719 cerr << "changed duration of phones with specified manner at "<<rate<<" percent\n";
00720 }
|
|
|
change duration of unstressed syllables. By changing duration of vowel in syllable (the shortening/lengthening of syllables effects primarily the vowels). The arg <both> decides whether word-accents are counted or not.
00650 {
00651 phon *pP;
00652 if (rate < MIN_SPEECHRATE || rate > MAX_SPEECHRATE)
00653 error ("phrase::changeDurNonStressed: rate < MIN_SPEECHRATE || rate > MAX_SPEECHRATE", -3);
00654
00655 if (both) {
00656 for (uint syl=0; syl<syllableVec.size(); syl++) {
00657 if (syllableVec[syl].Accent()==word ||syllableVec[syl].Accent()==none) {
00658 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00659 if (!(syl==0&&pho==0) &&
00660 !(syl==syllableVec.size()-1&&pho==syllableVec[syl].PhonNum()-1)) {
00661 pP = &syllableVec[syl].PhonVec()[pho];
00662 if ((*pP).Type()==vowel) {
00663 uint newDur = (*pP).Dur() * rate/100;
00664 (*pP).changeDur(newDur, glob);
00665 }
00666 }
00667 }
00668 }
00669 }
00670 } else { // only non-accent
00671 for (uint syl=0; syl<syllableVec.size(); syl++) {
00672 if (syllableVec[syl].Accent()==none) {
00673 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00674 if (!(syl==0&&pho==0) &&
00675 !(syl==syllableVec.size()-1&&pho==syllableVec[syl].PhonNum()-1)) {
00676 pP = &syllableVec[syl].PhonVec()[pho];
00677 if ((*pP).Type()==vowel) {
00678 uint newDur = (*pP).Dur() * rate/100;
00679 (*pP).changeDur(newDur, glob);
00680 }
00681 }
00682 }
00683 }
00684 }
00685 }
00686 calcDur();
00687 if (glob.verbose) {
00688 if (both)
00689 cerr << "changed duration of non phrase-stressed syllables at "<<rate<<" percent\n";
00690 else
00691 cerr << "changed duration of non stressed syllables at "<<rate<<" percent\n";
00692 }
00693 }
|
|
|
change duration of stressed syllables. By changing duration of vowel in syllable (the shortening/lengthening of syllables effects primarily the vowels). The arg <both> decides whether word-accents are counted or not.
00592 {
00593 phon *pP;
00594 if (rate < MIN_SPEECHRATE || rate > MAX_SPEECHRATE)
00595 error ("phrase::changeDurStressed: rate < MIN_SPEECHRATE || rate > MAX_SPEECHRATE", -3);
00596
00597 if (both) {
00598 for (uint syl=0; syl<syllableVec.size(); syl++) {
00599 if (syllableVec[syl].Accent()==word ||syllableVec[syl].Accent()==sentence) {
00600 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00601 if (!(syl==0&&pho==0) &&
00602 !(syl==syllableVec.size()-1&&pho==syllableVec[syl].PhonNum()-1)) {
00603 pP = &syllableVec[syl].PhonVec()[pho];
00604 if ((*pP).Type()==vowel) {
00605 if (!(rate>100 && (*pP).Diphtong())) {
00606 uint newDur = (*pP).Dur() * rate/100;
00607 (*pP).changeDur(newDur, glob);
00608 }
00609 }
00610 }
00611 }
00612 }
00613 }
00614 } else { // only sentence-accent
00615 for (uint syl=0; syl<syllableVec.size(); syl++) {
00616 if (syllableVec[syl].Accent()==sentence) {
00617 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00618 if (!(syl==0&&pho==0) &&
00619 !(syl==syllableVec.size()-1&&pho==syllableVec[syl].PhonNum()-1)) {
00620
00621 pP = &syllableVec[syl].PhonVec()[pho];
00622 if ((*pP).Type()==vowel) {
00623 uint newDur = (*pP).Dur() * rate/100;
00624 (*pP).changeDur(newDur, glob);
00625 }
00626 }
00627 }
00628 }
00629 }
00630 }
00631 calcDur();
00632 if (glob.verbose) {
00633 if (both)
00634 cerr << "changed duration of all stressed syllables at "<<rate<<" percent\n";
00635 else
00636 cerr << "changed duration of phrase-stressed syllables at "<<rate<<" percent\n";
00637 }
00638 }
|
|
|
changes the F0-range of phrase. The f0-contour gets expanded or compressed with the mean-f0 of the last syllable as reference. That's fine with declaration sentences.
00864 {
00865 uint lastMean = syllableVec[syllableVec.size()-1].MeanF0();
00866 for (uint syl=0; syl<syllableVec.size(); syl++)
00867 syllableVec[syl].changeF0Range(rate, lastMean);
00868
00869 if (glob.verbose)
00870 cerr << "changed F0-range at "<<rate<<" percent\n";
00871 calcMeanF0();
00872 }
|
|
|
changes pitch-contour of last syllable to <contour>. <contour> can be straight, rise or fall
00900 {
00901 phon *pP;
00902 uint start=0, end;
00903 bool found=false;
00904 uint lastSyl=syllableVec.size()-1;
00905 switch (contour) {
00906 case straight:
00907 start=syllableVec[lastSyl].MeanF0();
00908 for (uint pho=0; pho<syllableVec[lastSyl].PhonNum(); pho++) {
00909 pP = &syllableVec[lastSyl].PhonVec()[pho];
00910 if ((*pP).Voiced()) {
00911 (*pP).interpolateF0Vec(start, start);
00912 }
00913 }
00914 break;
00915 case rise:
00916 found=false;
00917 for (uint pho=0; pho<syllableVec[lastSyl].PhonNum(); pho++) {
00918 pP = &syllableVec[lastSyl].PhonVec()[pho];
00919 for (uint i=0; i<(*pP).F0Vec().size(); i++) {
00920 if ((*pP).F0Vec()[i] != 0) {
00921 start = (*pP).F0Vec()[i];
00922 found=true;
00923 break;
00924 }
00925 }
00926 if (found)
00927 break;
00928 }
00929 for (uint pho=0; pho<syllableVec[lastSyl].PhonNum(); pho++) {
00930 pP = &syllableVec[lastSyl].PhonVec()[pho];
00931 if ((*pP).Voiced()) {
00932 end = (uint) gradient2Hz(start, (*pP).Dur(), gradient);
00933 // cout <<"start: "<< start << " " << end << "\n";
00934 (*pP).interpolateF0Vec(start, end);
00935 start = end;
00936 }
00937 }
00938 break;
00939 case fall:
00940 for (uint pho=0; pho<syllableVec[lastSyl].PhonNum(); pho++) {
00941 pP = &syllableVec[lastSyl].PhonVec()[pho];
00942 if ((*pP).MeanF0()!=0) {
00943 start=(*pP).MeanF0();
00944 break;
00945 }
00946 }
00947 for (uint pho=0; pho<syllableVec[lastSyl].PhonNum(); pho++) {
00948 pP = &syllableVec[lastSyl].PhonVec()[pho];
00949 if ((*pP).Voiced()) {
00950 end = start - ((uint) gradient2Hz(start, (*pP).Dur(), gradient) - start);
00951 // cout <<"start: "<< start << " " << end << "\n";
00952 (*pP).interpolateF0Vec(start, end);
00953 start = end;
00954 }
00955 }
00956 break;
00957 }
00958 syllableVec[lastSyl].calcMeanF0();
00959 calcMeanF0();
00960 if (glob.verbose)
00961 cerr << "changed f0-countour of last syllable to gradient "<<gradient<<" \n";
00962 }
|
|
|
all nonzero f0Vals get multiplied by <rate> in percent. (so rate=100 means no change).
00766 {
00767 for (uint syl=0; syl<syllableVec.size(); syl++) {
00768 syllableVec[syl].changeMeanF0(rate);
00769 }
00770 calcMeanF0();
00771 if (glob.verbose)
00772 cerr << "changed avr. F0 at "<<rate<<" percent\n";
00773 }
|
|
|
changes pitch-contour of whole phrase to <contour>. <contour> can be straight, rise or fall
|
|
|
changes pitch-contour of regions of phrase divided by phraseaccents to <contour>. Like changePhraseContour(), but repeated for every accented syllable. A straight contour is not implemented, if rate>100 it's a rise and if rate<100 it's a fall.
01150 {
01151 uint sylC=0, tmpRate;
01152 for (uint syl=0; syl<syllableVec.size(); syl++) {
01153 if (syllableVec[syl].Accent()==sentence) { // if the syl has accent
01154 int step = ((int)rate - 100)/(int)sylC;
01155 tmpRate=0;
01156 for (uint tmpSyl=syl-sylC; tmpSyl <= syl; tmpSyl++) { // foreach syl before the accent
01157 tmpRate = 100 + (step * (tmpSyl-(syl-sylC)));
01158 // cout << tmpSyl << " " << step<<" "<<tmpRate << " " << rate<<"\n";
01159 syllableVec[tmpSyl].changeMeanF0(tmpRate);
01160
01161 }
01162 sylC=0;
01163 } else
01164 sylC++;
01165 }
01166 if (glob.verbose)
01167 cerr << "changed f0-countour to accented syllables at rate "<<rate<<" \n";
01168 calcMeanF0();
01169 }
|
|
|
changes pitch-contour of stressed syllables to <contour>. <contour> can be straight, rise or fall
00968 {
00969 phon *pP;
00970 uint start=0, end;
00971 bool found=false;
00972 if (both) {
00973 for (uint syl=0; syl<syllableVec.size()-1; syl++) {
00974 found=false;
00975 start=0;
00976 if (syllableVec[syl].Accent()==word || syllableVec[syl].Accent()==sentence) {
00977 // cout << syllableVec[syl].MeanF0() << "\n";
00978
00979 switch (contour) {
00980 case straight:
00981 syllableVec[syl].changeF0Range(gradient);
00982 break;
00983 case rise:
00984 found=false;
00985 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00986 pP = &syllableVec[syl].PhonVec()[pho];
00987 for (uint i=0; i<(*pP).F0Vec().size(); i++) {
00988 if ((*pP).F0Vec()[i] != 0) {
00989 start = (*pP).F0Vec()[i];
00990 found=true;
00991 break;
00992 }
00993 }
00994 if (found)
00995 break;
00996 }
00997 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00998 pP = &syllableVec[syl].PhonVec()[pho];
00999 if ((*pP).Voiced()) {
01000 end = (uint) gradient2Hz(start, (*pP).Dur(), gradient);
01001 // cout <<"start: "<< start << " " << end << "\n";
01002 (*pP).interpolateF0Vec(start, end);
01003 start = end;
01004 }
01005 }
01006 break;
01007 case fall:
01008 start = syllableVec[syl].MeanF0();
01009 // start fall from mean f0Val
01010 /*
01011 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01012 pP = &syllableVec[syl].PhonVec()[pho];
01013 if ((*pP).MeanF0()!=0) {
01014 start=(*pP).MeanF0();
01015 break;
01016 }
01017 }
01018 */
01019 // start fall from first nonzero f0Val
01020 // for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01021 // pP = &syllableVec[syl].PhonVec()[pho];
01022 // for (uint i=0; i<(*pP).FrameNum(); i++) {
01023 // if ((*pP).F0Vec()[i]!=0) {
01024 //start=(*pP).F0Vec()[i];
01025 //break;
01026 // }
01027 //}
01028 //}
01029 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01030 pP = &syllableVec[syl].PhonVec()[pho];
01031 if ((*pP).Voiced()) {
01032 end = start - ((uint) gradient2Hz(start, (*pP).Dur(), gradient) - start);
01033 // cout <<"start: "<< start << " " << end << "\n";
01034 (*pP).interpolateF0Vec(start, end);
01035 start = end;
01036 }
01037 }
01038 break;
01039 }
01040 }
01041 }
01042 } else { // only sentence-accent
01043 for (uint syl=0; syl<syllableVec.size()-1; syl++) {
01044 if (syllableVec[syl].Accent()==sentence) {
01045 switch (contour) {
01046 case straight:
01047 syllableVec[syl].changeF0Range(gradient);
01048 break;
01049 case rise:
01050 found=false;
01051 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01052 pP = &syllableVec[syl].PhonVec()[pho];
01053 for (uint i=0; i<(*pP).F0Vec().size(); i++) {
01054 if ((*pP).F0Vec()[i] != 0) {
01055 start = (*pP).F0Vec()[i];
01056 found=true;
01057 break;
01058 }
01059 }
01060 if (found)
01061 break;
01062 }
01063 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01064 pP = &syllableVec[syl].PhonVec()[pho];
01065 if ((*pP).Voiced()) {
01066 end = (uint) gradient2Hz(start, (*pP).Dur(), gradient);
01067 (*pP).interpolateF0Vec(start, end);
01068 start = end;
01069 }
01070 }
01071 break;
01072 case fall:
01073 // find first nonzero f0Val
01074 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01075 pP = &syllableVec[syl].PhonVec()[pho];
01076 for (uint i=0; i<(*pP).FrameNum(); i++) {
01077 if ((*pP).F0Vec()[i]!=0) {
01078 start=(*pP).F0Vec()[i];
01079 break;
01080 }
01081 }
01082 }
01083 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01084 pP = &syllableVec[syl].PhonVec()[pho];
01085 if ((*pP).Voiced()) {
01086 end = start - ((uint) gradient2Hz(start, (*pP).Dur(), gradient) - start);
01087 (*pP).interpolateF0Vec(start, end);
01088 start = end;
01089 }
01090 }
01091 break;
01092 }
01093 }
01094 }
01095 }
01096 if (glob.verbose)
01097 cerr << "changed f0-countour of stressed syllables to gradient "<<gradient<<" \n";
01098 calcMeanF0();
01099 }
|
|
|
change avr pitch of the syllables before stressed syllables. All nonzero f0Vals get multiplied by <rate> in percent. The arg <both> decides whether word-accents are counted or not. The last syllable is not affected (even if stressed), but has to be provided specially.
00836 {
00837 if (both) {
00838 for (uint syl=0; syl<syllableVec.size()-1; syl++) {
00839 if (syllableVec[syl+1].Accent()==word ||syllableVec[syl+1].Accent()==sentence) {
00840 syllableVec[syl].changeMeanF0(rate);
00841 }
00842 }
00843 } else { // only sentence-accent
00844 for (uint syl=0; syl<syllableVec.size(); syl++) {
00845 if (syllableVec[syl+1].Accent()==sentence) {
00846 syllableVec[syl].changeMeanF0(rate);
00847 }
00848 }
00849 }
00850 if (both)
00851 cerr << "changed avr. F0 of all syls prior to stress at "<<rate<<" percent\n";
00852 else
00853 cerr << "changed avr. F0 of syls prior to phrase-stress at "<<rate<<" percent\n";
00854 calcMeanF0();
00855 }
|
|
|
changes duration of all phones. if method = 0, then change duration of all phones in the utterance by factor <rate> if method = 1, then phrasestressed vowels get 100% change, wordstressed vowels get 90 % change, all others get 80 % change (phonetically adequate)
00526 {
00527 int absChange, wordChange, vowelChange, consChange;
00528 phon *pP;
00529 uint newDur;
00530 if (method == 0) {
00531 for (uint syl=0; syl<syllableVec.size(); syl++) {
00532 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00533 if (!(syl==0&&pho==0) &&
00534 !(syl==syllableVec.size()-1&&pho==syllableVec[syl].PhonNum()-1)) {
00535
00536 pP = &syllableVec[syl].PhonVec()[pho];
00537 if (!(rate>100 && syllableVec[syl].PhonVec()[pho].Diphtong())) {
00538 newDur = (*pP).Dur() * rate/100;
00539 (*pP).changeDur(newDur, glob);
00540 }
00541 }
00542 }
00543 }
00544 } else if (method == 1) {
00545 absChange = abs(100-(int)rate);
00546 if (rate>100) {
00547 wordChange = 100 + (int) (absChange*0.9);
00548 vowelChange = 100 + (int) (absChange*0.7);
00549 consChange = 100 + (int) (absChange*0.5);
00550 } else {
00551 wordChange = 100 - (int) (absChange*0.9);
00552 vowelChange = 100 - (int) (absChange*0.7);
00553 consChange = 100 - (int) (absChange*0.5);
00554 }
00555 // cout << wordChange<<" "<<vowelChange<<" "<<consChange<<"\n";
00556 for (uint syl=0; syl<syllableVec.size(); syl++) {
00557 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00558 if (!(syl==0&&pho==0) &&
00559 !(syl==syllableVec.size()-1&&pho==syllableVec[syl].PhonNum()-1)) {
00560
00561 pP = &syllableVec[syl].PhonVec()[pho];
00562 if (syllableVec[syl].Accent()==sentence && (*pP).Type()==vowel)
00563 newDur = syllableVec[syl].PhonVec()[pho].Dur() * rate/100;
00564 else if (syllableVec[syl].Accent()==word && (*pP).Type()==vowel)
00565 newDur = syllableVec[syl].PhonVec()[pho].Dur() * wordChange/100;
00566 else if ((*pP).Type()==vowel)
00567 newDur = syllableVec[syl].PhonVec()[pho].Dur() * vowelChange/100;
00568 else
00569 newDur = syllableVec[syl].PhonVec()[pho].Dur() * consChange/100;
00570
00571 syllableVec[syl].PhonVec()[pho].changeDur(newDur, glob);
00572 }
00573 }
00574 }
00575 } else {
00576 error("phrase::changeSpeechRate: undefined method", -3);
00577 }
00578 calcDur();
00579 if (glob.verbose)
00580 cerr << "changed speechrate at "<<rate<<" percent\n";
00581 }
|
|
|
change mean pitch of stressed syllables. All nonzero f0Vals get multiplied by <rate> in percent. The arg <both> decides whether word-accents are counted or not. The last syllable is not affected (even if stressed), but has to be provided specially.
00802 {
00803 if (both) {
00804 for (uint syl=0; syl<syllableVec.size()-1; syl++) {
00805 if (syllableVec[syl].Accent()==word || syllableVec[syl].Accent()==sentence) {
00806 // cout << syllableVec[syl].MeanF0() << " ";
00807 syllableVec[syl].changeMeanF0(rate);
00808 // cout << syllableVec[syl].MeanF0() << "\n";
00809
00810 }
00811 }
00812 } else { // only sentence-accent
00813 for (uint syl=0; syl<syllableVec.size(); syl++) {
00814 if (syllableVec[syl].Accent()==sentence) {
00815 syllableVec[syl].changeMeanF0(rate);
00816 }
00817 }
00818 }
00819 if (both)
00820 cerr << "changed avr. F0 of all stressed syllables at "<<rate<<" percent\n";
00821 else
00822 cerr << "changed avr. F0 of phrase-stressed syllables at "<<rate<<" percent\n";
00823 calcMeanF0();
00824 }
|
|
|
change intensity of stressed syllables. The arg <both> decides whether word-accents are counted or not. The dBval is added to intensity of the vowels in the syllables, so it can be negative to lower the intensity. If the result'd become negative, it's set to zero.
00732 {
00733 if (both) {
00734 for (uint syl=0; syl<syllableVec.size()-1; syl++) {
00735 if (syllableVec[syl].Accent()==sentence) {
00736 syllableVec[syl].changeAvrInt(dBVal);
00737 }
00738 if (syllableVec[syl].Accent()==word) {
00739 syllableVec[syl].changeAvrInt(dBVal/2);
00740 }
00741
00742 }
00743 } else { // only sentence-accent
00744 for (uint syl=0; syl<syllableVec.size(); syl++) {
00745 if (syllableVec[syl].Accent()==sentence) {
00746 syllableVec[syl].changeAvrInt(dBVal);
00747 }
00748 }
00749 }
00750 if (both)
00751 cerr << "changed avr. intensuty of all stressed syllables at "<<dBVal<<" dB\n";
00752 else
00753 cerr << "changed avr. intensity of phrase-stressed syllables at "<<dBVal<<" dB\n";
00754 }
|
|
|
changes the F0-variability. The f0-contour of each syllable gets expanded or compressed with the mean-f0 of the syllable as reference.
00881 {
00882 for (uint syl=0; syl<syllableVec.size(); syl++) {
00883 syllableVec[syl].changeF0Range(rate, syllableVec[syl].MeanF0());
00884 }
00885 if (glob.verbose)
00886 cerr << "changed F0-variability at "<<rate<<" percent\n";
00887 calcMeanF0();
00888 }
|
|
|
model coarticulation effects. Each phone gets features added at the borders from the neighboring phones, depending on coType:
01421 {
01422 phon *pP;
01423 for (uint syl=0; syl<syllableVec.size(); syl++) {
01424 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01425 pP = &syllableVec[syl].PhonVec()[pho];
01426 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01427 tmpPhonFeatures.coarticulation = true;
01428 tmpPhonFeatures.coarticulationRate = rate;
01429 tmpPhonFeatures.coarticulationType = coType;
01430 (*pP).setPhonFeatures(tmpPhonFeatures);
01431 }
01432 }
01433 if(glob.verbose)
01434 cerr << "added coarticulation with rate: "<<rate<<"\n";
01435 }
|
|
|
experiment: shift formant-values to model different head-sizes. not really implemented yet.
01396 {
01397 phon *pP;
01398 for (uint syl=0; syl<syllableVec.size(); syl++) {
01399 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01400 pP = &syllableVec[syl].PhonVec()[pho];
01401 if ((*pP).Voiced()) {
01402 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01403 tmpPhonFeatures.headSize = true;
01404 tmpPhonFeatures.headSizeRate = rate;
01405 (*pP).setPhonFeatures(tmpPhonFeatures);
01406 }
01407 }
01408 }
01409 if (glob.verbose)
01410 cerr <<"changed head-size at rate: "<<rate<<"\n";
01411 }
|
|
|
fill a phrase with one syllable with one phone of duration dur (in msec) with parameters taken from globals. Useful for finding the right parameters for certain phones.
|
|
|
model voice-source and articulation-tract filter parameters for each phone according to it's assigned features. sets bool xxxDyn true: i.e. all changung parmaters are marked to be dynamic.
00145 {
00146 phon *pP, *nextpP, *prevpP=NULL;
00147
00148 for (uint syl=0; syl<syllableVec.size(); syl++) {
00149 uint sylMeanF0 = syllableVec[syl].MeanF0();
00150 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00151 if (pho==syllableVec[syl].PhonNum()-1)
00152 if (syl==syllableVec.size()-1)
00153 nextpP = NULL;
00154 else
00155 nextpP = &(syllableVec[syl+1].PhonVec()[0]);
00156 else
00157 nextpP = &(syllableVec[syl].PhonVec()[pho+1]);
00158
00159 pP = &(syllableVec[syl].PhonVec()[pho]);
00160 (*pP).modelFeatures(sylMeanF0, globals, prevpP, nextpP);
00161 prevpP = pP;
00162 }
00163 }
00164 }
|
|
|
model voice-source and articulation-tract filter parameters for each phone for normal voice. They can later be overwritten by modelFeatures. sideeffect: sets bool xxxDyn true: i.e. all changung parmaters are marked to be dynamic.
00134 {
00135 phon *tmpPhonP;
00136 for (uint syl=0; syl<syllableVec.size(); syl++) {
00137 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00138 tmpPhonP = &(syllableVec[syl].PhonVec()[pho]);
00139 (*tmpPhonP).setSoundSource(globals);
00140 (*tmpPhonP).setArticulationTractFilter(globals);
00141 }
00142 }
00143 }
|
|
|
enhance the vowel precision. All formantVals in f1 and f2 of vowels in syllables with accent get displaced away from the center freqencies by adding the percentage by rate and distance to center frequency.
01327 {
01328 phon *pP;
01329 for (uint syl=0; syl<syllableVec.size(); syl++) {
01330 if (syllableVec[syl].Accent()==accent) {
01331 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01332 pP = &syllableVec[syl].PhonVec()[pho];
01333 if ((*pP).Type()==vowel) { // && ! (*pP).Diphtong()) {
01334 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01335 tmpPhonFeatures.overshoot = true;
01336 tmpPhonFeatures.overshootRate = rate;
01337 (*pP).setPhonFeatures(tmpPhonFeatures);
01338 }
01339 }
01340 }
01341 }
01342 if (glob.verbose)
01343 cerr << "added overshoot at "<<rate<<" for accent-type: "<<(uint)accent<<"\n";
01344 }
|
|
|
prints klatt-file (iles&simmons-format). If used, the bool sensyn in the global paramters should be false.
00121 {
00122 outFile << "# parameter file for the klatt-formant speech-synthesizer, implementation of iles & simmons, (version with comments possible!) \n";
00123 outFile << "# generated by emoSyn - speechsynthesizer-frontend for emotional speech, author felixbur@kgw.tu-berlin.de\n";
00124 for (uint syl=0; syl<syllableVec.size(); syl++) {
00125 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00126 outFile << "# name of phone: " << syllableVec[syl].PhonVec()[pho].Name() << "\n";
00127 outFile << "# meaning of parameters: \n";
00128 outFile << "# f0 av f1 b1 f2 b2 f3 b3 f4 b4 f5 b5 f6 b6 fnz bnz fnp bnp asp kopen aturb tilt af skew a1 b1p a2 b2p a3 b3p a4 b4p a5 b5p a6 b6p anp ab avp gain\n# values:\n";
00129 syllableVec[syl].PhonVec()[pho].printKlattFrames(outFile);
00130 }
00131 }
00132 }
|
|
|
prints label-file (esps-format).
00200 {
00201 double time=0;
00202 phon *pP;
00203
00204 outFile << "signal emoSyn-output\n";
00205 outFile << "type 0\n";
00206 outFile << "comment created by emoSyn\n";
00207 outFile << "font -misc-*-bold-*-*-*-15-*-*-*-*-*-*-*\n";
00208 outFile << "separator ;\n";
00209 outFile << "nfields 1\n";
00210 outFile << "#\n";
00211 for (uint syl=0; syl<syllableVec.size(); syl++) {
00212 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00213 pP = &syllableVec[syl].PhonVec()[pho];
00214 outFile << "\t" << time << "\t-1 ";
00215 outFile << (*pP).Name();
00216 outFile << "\n";
00217 // print features: overshoot, undershoot
00218 if ((*pP).PhonFeatures().overshoot) {
00219 outFile << "\t" << time << "\t-1 " << "+overshoot\n";
00220 }
00221 if ((*pP).PhonFeatures().undershoot) {
00222 outFile << "\t" << time << "\t-1 " << "+undershoot\n";
00223 }
00224 // print features: breathy
00225 if ((*pP).PhonFeatures().breathy) {
00226 double bS = (*pP).PhonFeatures().brtStart*(double) globals.ui / 1000;
00227 outFile << "\t" << time + bS << "\t-1 " << "+brt\n";
00228 if ((*pP).PhonFeatures().brtStop < (*pP).Dur()) {
00229 double bE = (*pP).PhonFeatures().brtStop*(double) globals.ui / 1000;
00230 outFile << "\t" << time + bE << "\t-1 " << "-brt\n";
00231 }
00232 }
00233 // print features: voiced
00234 if ((*pP).PhonFeatures().voiced) {
00235 if (!(*pP).Voiced()) {
00236 double vS = (*pP).PhonFeatures().vocStart*(double) globals.ui / 1000;
00237 outFile << "\t" << time + vS << "\t-1 " << "+voc\n";
00238 if ((*pP).PhonFeatures().vocStop < (*pP).Dur()) {
00239 double vE = (*pP).PhonFeatures().vocStop*(double) globals.ui / 1000;
00240 outFile << "\t" << time + vE << "\t-1 " << "-voc\n";
00241 }
00242 } else {
00243 double vE = (*pP).PhonFeatures().vocStop*(double) globals.ui / 1000;
00244 outFile << "\t" << time + vE << "\t-1 " << "-voc\n";
00245 if ((*pP).PhonFeatures().vocStart < (*pP).Dur()) {
00246 double vS = (*pP).PhonFeatures().vocStart*(double) globals.ui / 1000;
00247 outFile << "\t" << time + vS << "\t-1 " << "+voc\n";
00248 }
00249 }
00250 }
00251 // print features: laryngealized
00252 if ((*pP).PhonFeatures().creaky) {
00253 double bS = (*pP).PhonFeatures().creakyStart*(double) globals.ui / 1000;
00254 outFile << "\t" << time + bS << "\t-1 " << "+lar\n";
00255 if ((*pP).PhonFeatures().creakyStop < (*pP).Dur()) {
00256 double bE = (*pP).PhonFeatures().creakyStop*(double) globals.ui / 1000;
00257 outFile << "\t" << time + bE << "\t-1 " << "-lar\n";
00258 }
00259 }
00260 time += (double) (*pP).Dur()/1000;
00261 }
00262 }
00263 outFile << "\n";
00264 }
|
|
|
prints pho-file (mbrola-format). It's used for resynthesis. If the global parameter copyDB is true, the syllable borders get printed as well.
00166 {
00167 vector<uint> tmpF0ValVec(0);
00168 vector<uint> tmpF0TimeVec(0);
00169 phon *pP;
00170 uint time;
00171 for (uint syl=0; syl<syllableVec.size(); syl++) {
00172 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00173 pP = &syllableVec[syl].PhonVec()[pho];
00174 if (glob.copyDB) {
00175 outFile << (*pP).Name() << " ";
00176 outFile << (*pP).Dur() << " ";
00177 if ((*pP).Voiced()) {
00178 for (uint f0Val=0; f0Val<(*pP).F0Vec().size();f0Val++) {
00179 time = f0Val*(100/(*pP).F0Vec().size());
00180 outFile << time << " " << (*pP).F0Vec()[f0Val]/10 << " ";
00181 }
00182 }
00183 } else {
00184 tmpF0ValVec = (*pP).F0ValVec();
00185 tmpF0TimeVec = (*pP).F0TimeVec();
00186 outFile << (*pP).Name() << " ";
00187 outFile << (*pP).Dur() << " ";
00188 for (uint f0Val=0; f0Val<(*pP).FrameNum();f0Val++)
00189 outFile << tmpF0TimeVec[f0Val] << " "
00190 << tmpF0ValVec[f0Val]*10 << " ";
00191 }
00192 outFile << "\n";
00193 }
00194 if (glob.copyDB)
00195 outFile << "- " << (int) syllableVec[syl].Accent() << "\n";
00196 }
00197 outFile << "\n";
00198 }
|
|
|
prints phonfeatures to outFile (debugging).
00089 {
00090 phon *pP;
00091 for (uint syl=0; syl<syllableVec.size(); syl++) {
00092 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00093 pP = &syllableVec[syl].PhonVec()[pho];
00094 outFile << (*pP).Name() << " " << (*pP).Diphtong() << "\n";
00095 //tmpPhonFeatures.voiced << "\n";
00096 }
00097 }
00098 }
|
|
|
prints means of f0-contour per phon to outFile (debugging).
00101 {
00102 vector<phon> tmpPhonVec;
00103 for (uint syl=0; syl<syllableVec.size(); syl++) {
00104 tmpPhonVec = syllableVec[syl].PhonVec();
00105 for (uint pho=0; pho<tmpPhonVec.size(); pho++) {
00106 outFile << tmpPhonVec[pho].MeanF0() << " ";
00107 }
00108 outFile << "\t";
00109 }
00110 outFile << "\n";
00111 }
|
|
|
prints klatt-file (sensymetrics-format). If used, the bool sensyn in the global paramters should be true.
01667 {
01668 uint start=0;
01669 outFile << "parameter file for the klatt-formant speech-synthesizer\n implementation: sensyn, sensimetrics coorp. \n";
01670 outFile << "generated by emoSyn\n - speechsynthesizer-frontend for emotional speech,\n author felixbur@globw.tu-berlin.de\n";
01671 outFile << "6\n7\n8\n9\n";
01672 outFile << "60 parameters\n";
01673 outFile << "11\n12\n13\n";
01674
01675 outFile << "DU C 30 "<<dur<<" 5000 Duration of the utterance, in msec\n";
01676 outFile << "UI C 1 "<<glob.ui<<" 20 Update interval for parameter reset, in msec\n";
01677 outFile << "SR C 5000 "<<glob.sr<<" 20000 Output sampling rate, in samples/sec\n";
01678 outFile << "NF C 1 "<<glob.nf<<" 6 Number of formants in cascade branch\n";
01679 outFile << "SS C 1 "<<glob.ss<<" 3 Source switch (1=impulse, 2=natural, 3=LF model)\n";
01680 outFile << "RS C 1 "<<glob.rs<<" 8191 Random seed (initial value of random # generator)\n";
01681 outFile << "SB C 0 "<<glob.sb<<" 1 Same noise burst, reset RS if AF=AH=0, 0=no,1=yes\n";
01682 outFile << "CP C 0 "<<glob.cp<<" 1 0=Cascade, 1=Parallel tract excitation by AV\n";
01683 outFile << "OS C 0 "<<glob.os<<" 20 Output selector (0=normal,1=voicing source,...)\n";
01684 outFile << "GV C 0 "<<glob.gv<<" 80 Overall gain scale factor for AV, in dB\n";
01685 outFile << "GH C 0 "<<glob.gh<<" 80 Overall gain scale factor for AH, in dB\n";
01686 outFile << "GF C 0 "<<glob.gf<<" 80 Overall gain scale factor for AF, in dB\n";
01687 if (glob.f0Dyn)
01688 outFile << "F0 V 0 "<<glob.f0<<" 5000 Fundamental frequency, in tenths of a Hz\n";
01689 else
01690 outFile << "F0 v 0 "<<glob.f0<<" 5000 Fundamental frequency, in tenths of a Hz\n";
01691 if (glob.avDyn)
01692 outFile << "AV V 0 "<<glob.av<<" 80 Amplitude of voicing, in dB\n";
01693 else
01694 outFile << "AV v 0 "<<glob.av<<" 80 Amplitude of voicing, in dB\n";
01695 if (glob.oqDyn)
01696 outFile << "OQ V 10 "<<glob.oq<<" 99 Open quotient (voicing open-time/period), in %\n";
01697 else
01698 outFile << "OQ v 10 "<<glob.oq<<" 99 Open quotient (voicing open-time/period), in %\n";
01699 if (glob.sqDyn)
01700 outFile << "SQ V 100 "<<glob.sq<<" 500 Speed quotient (rise/fall time, LF model), in %\n";
01701 else
01702 outFile << "SQ v 100 "<<glob.sq<<" 500 Speed quotient (rise/fall time, LF model), in %\n";
01703 if (glob.tlDyn)
01704 outFile << "TL V 0 "<<glob.tl<<" 41 Extra tilt of voicing spectrum, dB down @ 3 kHz\n";
01705 else
01706 outFile << "TL v 0 "<<glob.tl<<" 41 Extra tilt of voicing spectrum, dB down @ 3 kHz\n";
01707 if (glob.flDyn)
01708 outFile << "FL V 0 "<<glob.fl<<" 100 Flutter (random fluct in f0), in % of maximum\n";
01709 else
01710 outFile << "FL v 0 "<<glob.fl<<" 100 Flutter (random fluct in f0), in % of maximum\n";
01711 if (glob.diDyn)
01712 outFile << "DI V 0 "<<glob.di<<" 100 Diplophonia (alt periods closer), in % of max\n";
01713 else
01714 outFile << "DI v 0 "<<glob.di<<" 100 Diplophonia (alt periods closer), in % of max\n";
01715 if (glob.ahDyn)
01716 outFile << "AH V 0 "<<glob.ah<<" 80 Amplitude of aspiration, in dB\n";
01717 else
01718 outFile << "AH v 0 "<<glob.ah<<" 80 Amplitude of aspiration, in dB\n";
01719 if (glob.afDyn)
01720 outFile << "AF V 0 "<<glob.af<<" 80 Amplitude of frication, in dB\n";
01721 else
01722 outFile << "AF v 0 "<<glob.af<<" 80 Amplitude of frication, in dB\n";
01723 if (glob.f1Dyn)
01724 outFile << "F1 V 180 "<<glob.f1<<" 1300 Frequency of 1st formant, in Hz\n";
01725 else
01726 outFile << "F1 v 180 "<<glob.f1<<" 1300 Frequency of 1st formant, in Hz\n";
01727 if (glob.b1Dyn)
01728 outFile << "B1 V 30 "<<glob.b1<<" 1000 Bandwidth of 1st formant, in Hz\n";
01729 else
01730 outFile << "B1 v 30 "<<glob.b1<<" 1000 Bandwidth of 1st formant, in Hz\n";
01731 if (glob.df1Dyn)
01732 outFile << "DF1 V 0 "<<glob.df1<<" 100 Change in F1 during open portion of period, in Hz\n";
01733 else
01734 outFile << "DF1 v 0 "<<glob.df1<<" 100 Change in F1 during open portion of period, in Hz\n";
01735 if (glob.db1Dyn)
01736 outFile << "DB1 V 0 "<<glob.db1<<" 400 Change in B1 during open portion of period, in Hz\n";
01737 else
01738 outFile << "DB1 v 0 "<<glob.db1<<" 400 Change in B1 during open portion of period, in Hz\n";
01739 if (glob.f2Dyn)
01740 outFile << "F2 V 550 "<<glob.f2<<" 3000 Frequency of 2nd formant, in Hz\n";
01741 else
01742 outFile << "F2 v 550 "<<glob.f2<<" 3000 Frequency of 2nd formant, in Hz\n";
01743 if (glob.b2Dyn)
01744 outFile << "B2 V 40 "<<glob.b2<<" 90 Bandwidth of 2nd formant, in Hz\n";
01745 else
01746 outFile << "B2 v 40 "<<glob.b2<<" 90 Bandwidth of 2nd formant, in Hz\n";
01747 if (glob.f3Dyn)
01748 outFile << "F3 V 1200 "<<glob.f3<<" 2500 Frequency of 3rd formant, in Hz\n";
01749 else
01750 outFile << "F3 v 1200 "<<glob.f3<<" 2500 Frequency of 3rd formant, in Hz\n";
01751 if (glob.b3Dyn)
01752 outFile << "B3 V 60 "<<glob.b3<<" 150 Bandwidth of 3rd formant, in Hz\n";
01753 else
01754 outFile << "B3 v 60 "<<glob.b3<<" 150 Bandwidth of 3rd formant, in Hz\n";
01755 if (glob.f4Dyn)
01756 outFile << "F4 V 2400 "<<glob.f4<<" 4990 Frequency of 4th formant, in Hz\n";
01757 else
01758 outFile << "F4 v 2400 "<<glob.f4<<" 4990 Frequency of 4th formant, in Hz\n";
01759 if (glob.b4Dyn)
01760 outFile << "B4 V 100 "<<glob.b4<<" 1000 Bandwidth of 4th formant, in Hz\n";
01761 else
01762 outFile << "B4 v 100 "<<glob.b4<<" 1000 Bandwidth of 4th formant, in Hz\n";
01763 if (glob.f5Dyn)
01764 outFile << "F5 V 3000 "<<glob.f5<<" 4990 Frequency of 5th formant, in Hz\n";
01765 else
01766 outFile << "F5 v 3000 "<<glob.f5<<" 4990 Frequency of 5th formant, in Hz\n";
01767 if (glob.b5Dyn)
01768 outFile << "B5 V 100 "<<glob.b5<<" 1500 Bandwidth of 5th formant, in Hz\n";
01769 else
01770 outFile << "B5 v 100 "<<glob.b5<<" 1500 Bandwidth of 5th formant, in Hz\n";
01771 if (glob.f6Dyn)
01772 outFile << "F6 V 3000 "<<glob.f6<<" 4990 Frequency of 6th formant, in Hz (applies if NF=6)\n";
01773 else
01774 outFile << "F6 v 3000 "<<glob.f6<<" 4990 Frequency of 6th formant, in Hz (applies if NF=6)\n";
01775 if (glob.b6Dyn)
01776 outFile << "B6 V 100 "<<glob.b6<<" 4000 Bandwidth of 6th formant, in Hz (applies if NF=6)\n";
01777 else
01778 outFile << "B6 v 100 "<<glob.b6<<" 4000 Bandwidth of 6th formant, in Hz (applies if NF=6)\n";
01779 if (glob.fnpDyn)
01780 outFile << "FNP V 180 "<<glob.fnp<<" 2000 Frequency of nasal pole, in Hz\n";
01781 else
01782 outFile << "FNP v 180 "<<glob.fnp<<" 2000 Frequency of nasal pole, in Hz\n";
01783 if (glob.bnpDyn)
01784 outFile << "BNP V 40 "<<glob.bnp<<" 1000 Bandwidth of nasal pole, in Hz\n";
01785 else
01786 outFile << "BNP v 40 "<<glob.bnp<<" 1000 Bandwidth of nasal pole, in Hz\n";
01787 if (glob.fnzDyn)
01788 outFile << "FNZ V 180 "<<glob.fnz<<" 2000 Frequency of nasal zero, in Hz\n";
01789 else
01790 outFile << "FNZ v 180 "<<glob.fnz<<" 2000 Frequency of nasal zero, in Hz\n";
01791 if (glob.bnzDyn)
01792 outFile << "BNZ V 40 "<<glob.bnz<<" 1000 Bandwidth of nasal zero, in Hz\n";
01793 else
01794 outFile << "BNZ v 40 "<<glob.bnz<<" 1000 Bandwidth of nasal zero, in Hz\n";
01795 if (glob.ftpDyn)
01796 outFile << "FTP V 300 "<<glob.ftp<<" 3000 Frequency of tracheal pole, in Hz\n";
01797 else
01798 outFile << "FTP v 300 "<<glob.ftp<<" 3000 Frequency of tracheal pole, in Hz\n";
01799 if (glob.btpDyn)
01800 outFile << "BTP V 40 "<<glob.btp<<" 1000 Bandwidth of tracheal pole, in Hz\n";
01801 else
01802 outFile << "BTP v 40 "<<glob.btp<<" 1000 Bandwidth of tracheal pole, in Hz\n";
01803 if (glob.ftzDyn)
01804 outFile << "FTZ V 300 "<<glob.ftz<<" 3000 Frequency of tracheal zero, in Hz\n";
01805 else
01806 outFile << "FTZ v 300 "<<glob.ftz<<" 3000 Frequency of tracheal zero, in Hz\n";
01807 if (glob.btzDyn)
01808 outFile << "BTZ V 40 "<<glob.btz<<" 2000 Bandwidth of tracheal zero, in Hz\n";
01809 else
01810 outFile << "BTZ v 40 "<<glob.btz<<" 2000 Bandwidth of tracheal zero, in Hz\n";
01811 if (glob.a2fDyn)
01812 outFile << "A2F V 0 "<<glob.a2f<<" 80 Amp of fric-excited parallel 2nd formant, in dB\n";
01813 else
01814 outFile << "A2F v 0 "<<glob.a2f<<" 80 Amp of fric-excited parallel 2nd formant, in dB\n";
01815 if (glob.a3fDyn)
01816 outFile << "A3F V 0 "<<glob.a3f<<" 80 Amp of fric-excited parallel 3rd formant, in dB\n";
01817 else
01818 outFile << "A3F v 0 "<<glob.a3f<<" 80 Amp of fric-excited parallel 3rd formant, in dB\n";
01819 if (glob.a4fDyn)
01820 outFile << "A4F V 0 "<<glob.a4f<<" 80 Amp of fric-excited parallel 4th formant, in dB\n";
01821 else
01822 outFile << "A4F v 0 "<<glob.a4f<<" 80 Amp of fric-excited parallel 4th formant, in dB\n";
01823 if (glob.a5fDyn)
01824 outFile << "A5F V 0 "<<glob.a5f<<" 80 Amp of fric-excited parallel 5th formant, in dB\n";
01825 else
01826 outFile << "A5F v 0 "<<glob.a5f<<" 80 Amp of fric-excited parallel 5th formant, in dB\n";
01827 if (glob.a6fDyn)
01828 outFile << "A6F V 0 "<<glob.a6f<<" 80 Amp of fric-excited parallel 6th formant, in dB\n";
01829 else
01830 outFile << "A6F v 0 "<<glob.a6f<<" 80 Amp of fric-excited parallel 6th formant, in dB\n";
01831 if (glob.abDyn)
01832 outFile << "AB V 0 "<<glob.ab<<" 80 Amp of fric-excited parallel bypass path, in dB\n";
01833 else
01834 outFile << "AB v 0 "<<glob.ab<<" 80 Amp of fric-excited parallel bypass path, in dB\n";
01835 if (glob.b2pDyn)
01836 outFile << "B2F V 40 "<<glob.b2p<<" 1000 Bw of fric-excited parallel 2nd formant, in Hz\n";
01837 else
01838 outFile << "B2F v 40 "<<glob.b2p<<" 1000 Bw of fric-excited parallel 2nd formant, in Hz\n";
01839 if (glob.b3pDyn)
01840 outFile << "B3F V 60 "<<glob.b3p<<" 1000 Bw of fric-excited parallel 3rd formant, in Hz\n";
01841 else
01842 outFile << "B3F v 60 "<<glob.b3p<<" 1000 Bw of fric-excited parallel 3rd formant, in Hz\n";
01843 if (glob.b4pDyn)
01844 outFile << "B4F V 100 "<<glob.b4p<<" 1000 Bw of fric-excited parallel 4th formant, in Hz\n";
01845 else
01846 outFile << "B4F v 100 "<<glob.b4p<<" 1000 Bw of fric-excited parallel 4th formant, in Hz\n";
01847 if (glob.b5pDyn)
01848 outFile << "B5F V 100 "<<glob.b5p<<" 1500 Bw of fric-excited parallel 5th formant, in Hz\n";
01849 else
01850 outFile << "B5F v 100 "<<glob.b5p<<" 1500 Bw of fric-excited parallel 5th formant, in Hz\n";
01851 if (glob.b6pDyn)
01852 outFile << "B6F V 100 "<<glob.b6p<<" 4000 Bw of fric-excited parallel 6th formant, in Hz\n";
01853 else
01854 outFile << "B6F v 100 "<<glob.b6p<<" 4000 Bw of fric-excited parallel 6th formant, in Hz\n";
01855 if (glob.anvDyn)
01856 outFile << "ANV V 0 "<<glob.anv<<" 80 Amp of voice-excited parallel nasal form., in dB\n";
01857 else
01858 outFile << "ANV v 0 "<<glob.anv<<" 80 Amp of voice-excited parallel nasal form., in dB\n";
01859 if (glob.a1vDyn)
01860 outFile << "A1V V 0 "<<glob.a1v<<" 80 Amp of voice-excited parallel 1st formant, in dB\n";
01861 else
01862 outFile << "A1V v 0 "<<glob.a1v<<" 80 Amp of voice-excited parallel 1st formant, in dB\n";
01863 if (glob.a2vDyn)
01864 outFile << "A2V V 0 "<<glob.a2v<<" 80 Amp of voice-excited parallel 2nd formant, in dB\n";
01865 else
01866 outFile << "A2V v 0 "<<glob.a2v<<" 80 Amp of voice-excited parallel 2nd formant, in dB\n";
01867 if (glob.a3vDyn)
01868 outFile << "A3V V 0 "<<glob.a3v<<" 80 Amp of voice-excited parallel 3rd formant, in dB\n";
01869 else
01870 outFile << "A3V v 0 "<<glob.a3v<<" 80 Amp of voice-excited parallel 3rd formant, in dB\n";
01871 if (glob.a4vDyn)
01872 outFile << "A4V V 0 "<<glob.a4v<<" 80 Amp of voice-excited parallel 4th formant, in dB\n";
01873 else
01874 outFile << "A4V v 0 "<<glob.a4v<<" 80 Amp of voice-excited parallel 4th formant, in dB\n";
01875 if (glob.atvDyn)
01876 outFile << "ATV V 0 "<<glob.atv<<" 80 Amp of voice-excited par tracheal formant, in dB\n";
01877 else
01878 outFile << "ATV v 0 "<<glob.atv<<" 80 Amp of voice-excited par tracheal formant, in dB\n";
01879
01880 outFile << "\n\n\n\n"; // skip four lines
01881
01882
01883 outFile << "Varied Parameters:\n";
01884 outFile.width(4);
01885 outFile << "time";
01886 if (glob.f0Dyn) outFile << " F0";
01887 if (glob.avDyn) outFile << " AV";
01888 if (glob.oqDyn) outFile << " OQ";
01889 if (glob.sqDyn) outFile << " SQ";
01890 if (glob.tlDyn) outFile << " TL";
01891 if (glob.flDyn) outFile << " FL";
01892 if (glob.diDyn) outFile << " DI";
01893 if (glob.ahDyn) outFile << " AH";
01894 if (glob.afDyn) outFile << " AF";
01895 if (glob.f1Dyn) outFile << " F1";
01896 if (glob.b1Dyn) outFile << " B1";
01897 if (glob.df1Dyn) outFile << " DF1";
01898 if (glob.db1Dyn) outFile << " DB1";
01899 if (glob.f2Dyn) outFile << " F2";
01900 if (glob.b2Dyn) outFile << " B2";
01901 if (glob.f3Dyn) outFile << " F3";
01902 if (glob.b3Dyn) outFile << " B3";
01903 if (glob.f4Dyn) outFile << " F4";
01904 if (glob.b4Dyn) outFile << " B4";
01905 if (glob.f5Dyn) outFile << " F5";
01906 if (glob.b5Dyn) outFile << " B5";
01907 if (glob.f6Dyn) outFile << " F6";
01908 if (glob.b6Dyn) outFile << " B6";
01909 if (glob.fnpDyn) outFile << " FNP";
01910 if (glob.bnpDyn) outFile << " BNP";
01911 if (glob.fnzDyn) outFile << " FNZ";
01912 if (glob.bnzDyn) outFile << " BNZ";
01913 if (glob.ftpDyn) outFile << " FTP";
01914 if (glob.btpDyn) outFile << " BTZ";
01915 if (glob.ftzDyn) outFile << " FTZ";
01916 if (glob.btzDyn) outFile << " BTZ";
01917 if (glob.a2fDyn) outFile << " A2F";
01918 if (glob.a3fDyn) outFile << " A3F";
01919 if (glob.a4fDyn) outFile << " A4F";
01920 if (glob.a5fDyn) outFile << " A5F";
01921 if (glob.a6fDyn) outFile << " A6F";
01922 if (glob.abDyn) outFile << " AB";
01923 if (glob.b2pDyn) outFile << " B2F";
01924 if (glob.b3pDyn) outFile << " B3F";
01925 if (glob.b4pDyn) outFile << " B4F";
01926 if (glob.b5pDyn) outFile << " B5F";
01927 if (glob.b6pDyn) outFile << " B6F";
01928 if (glob.anvDyn) outFile << " ANV";
01929 if (glob.a1vDyn) outFile << " A1V";
01930 if (glob.a2vDyn) outFile << " A2V";
01931 if (glob.a3vDyn) outFile << " A3V";
01932 if (glob.a4vDyn) outFile << " A4V";
01933 if (glob.atvDyn) outFile << " ATV";
01934
01935 outFile << "\n";
01936 for (uint syl=0; syl<syllableVec.size(); syl++) {
01937 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01938 outFile << "# name of phone: "<<syllableVec[syl].PhonVec()[pho].Name()<<"\n";
01939 outFile << "# meaning of params:\n";
01940 outFile << "# time";
01941 if (glob.f0Dyn) outFile << " F0";
01942 if (glob.avDyn) outFile << " AV";
01943 if (glob.oqDyn) outFile << " OQ";
01944 if (glob.sqDyn) outFile << " SQ";
01945 if (glob.tlDyn) outFile << " TL";
01946 if (glob.flDyn) outFile << " FL";
01947 if (glob.diDyn) outFile << " DI";
01948 if (glob.ahDyn) outFile << " AH";
01949 if (glob.afDyn) outFile << " AF";
01950 if (glob.f1Dyn) outFile << " F1";
01951 if (glob.b1Dyn) outFile << " B1";
01952 if (glob.df1Dyn) outFile << " DF1";
01953 if (glob.db1Dyn) outFile << " DB1";
01954 if (glob.f2Dyn) outFile << " F2";
01955 if (glob.b2Dyn) outFile << " B2";
01956 if (glob.f3Dyn) outFile << " F3";
01957 if (glob.b3Dyn) outFile << " B3";
01958 if (glob.f4Dyn) outFile << " F4";
01959 if (glob.b4Dyn) outFile << " B4";
01960 if (glob.f5Dyn) outFile << " F5";
01961 if (glob.b5Dyn) outFile << " B5";
01962 if (glob.f6Dyn) outFile << " F6";
01963 if (glob.b6Dyn) outFile << " B6";
01964 if (glob.fnpDyn) outFile << " FNP";
01965 if (glob.bnpDyn) outFile << " BNP";
01966 if (glob.fnzDyn) outFile << " FNZ";
01967 if (glob.bnzDyn) outFile << " BNZ";
01968 if (glob.ftpDyn) outFile << " FTP";
01969 if (glob.btpDyn) outFile << " BTZ";
01970 if (glob.ftzDyn) outFile << " FTZ";
01971 if (glob.btzDyn) outFile << " BTZ";
01972 if (glob.a2fDyn) outFile << " A2F";
01973 if (glob.a3fDyn) outFile << " A3F";
01974 if (glob.a4fDyn) outFile << " A4F";
01975 if (glob.a5fDyn) outFile << " A5F";
01976 if (glob.a6fDyn) outFile << " A6F";
01977 if (glob.abDyn) outFile << " AB";
01978 if (glob.b2pDyn) outFile << " B2F";
01979 if (glob.b3pDyn) outFile << " B3F";
01980 if (glob.b4pDyn) outFile << " B4F";
01981 if (glob.b5pDyn) outFile << " B5F";
01982 if (glob.b6pDyn) outFile << " B6F";
01983 if (glob.anvDyn) outFile << " ANV";
01984 if (glob.a1vDyn) outFile << " A1V";
01985 if (glob.a2vDyn) outFile << " A2V";
01986 if (glob.a3vDyn) outFile << " A3V";
01987 if (glob.a4vDyn) outFile << " A4V";
01988 if (glob.atvDyn) outFile << " ATV";
01989 outFile << "\n";
01990 syllableVec[syl].PhonVec()[pho].printSenSynFrames(start, outFile, glob);
01991 start += syllableVec[syl].PhonVec()[pho].FrameNum() * glob.ui;
01992 }
01993 }
01994 }
|
|
|
prints names of phones to outFile (debugging).
00078 {
00079 for (uint i=0; i<syllableVec.size(); i++) {
00080 syllableVec[i].printPhonVec(outFile);
00081 if (syllableVec[i].PhonNum()<4)
00082 outFile << "\t";
00083 outFile << "\taccentType = " << (uint) syllableVec[i].Accent()<<",";
00084 outFile << "\t";
00085 outFile << "avr. f0 = " << syllableVec[i].MeanF0() << "\n";
00086 }
00087 }
|
|
|
assignes only f0-vals. It's used for resynthesis if you only want to read in the original intonation contour to print it to a new input-file in connection with
02235 {
02236
02237 // variable declarations
02238 string actName, nextName, requiredDiphon;
02239 phon *actPhonP, *nextPhonP;
02240 string diphonType;
02241 string buffer; // tmp-buffer for line in diphon-file
02242 vector<string> bufferVec;
02243 PHON_MANNER actManner;
02244 uint lineCnt=0; // Cnt of lines in diphon-file
02245 vector<uint> diphonFrameNums(0); // list of frameNums foreach diphon
02246
02247 uint phonBrdFrm; // border frame foreach phon
02248 uint phonFrameNum; // frameNum in database of phon
02249 uint allFrameCnt=0; // help Cnt for assignment of params to phones
02250
02251 // vectors for markers in diphons
02252 vector<uint> t1(0), t2(0), border(0);
02253 // t1: first transition, t2: 2. transition
02254 // border between diphones
02255
02256 vector<uint> tmpF0;
02257
02258 uint phon=0;
02259 string diphonName;
02260 for (uint syl=0; syl<syllableVec.size(); syl++) {
02261 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
02262
02263 actPhonP = &syllableVec[syl].PhonVec()[pho];
02264 if (pho==syllableVec[syl].PhonNum()-1)
02265 if (syl==syllableVec.size()-1)
02266 break;
02267 else
02268 nextPhonP = &syllableVec[syl+1].PhonVec()[0];
02269 else
02270 nextPhonP = &syllableVec[syl].PhonVec()[pho+1];
02271
02272 actName = (*actPhonP).Name();
02273 nextName = (*nextPhonP).Name();
02274
02275 requiredDiphon = actName + "-" + nextName;
02276
02277 requiredDiphon = diphonBase + "/" + requiredDiphon + ".dip";
02278 // cout << requiredDiphon << "\n";
02279 ifstream inFile(requiredDiphon.c_str());
02280 if (inFile.good()) {
02281 lineCnt = 0;
02282 while (getline(inFile, buffer)) {
02283 if (buffer != "" && buffer.at(0) != '#') {
02284 bufferVec = string2vec(buffer);
02285 if (lineCnt==0) {
02286 diphonName = bufferVec[1];
02287 }
02288 if (lineCnt==1) {
02289 if (bufferVec[0]!="frameNumber") error("error in diphon-file:"+diphonName+"\n",-3);
02290 diphonFrameNums.push_back(atoi(bufferVec[1].c_str()));
02291 }
02292 else if (lineCnt==2) {
02293 if (bufferVec[0]!="t1") error("error in diphon-file:"+diphonName+"\n",-3);
02294 t1.push_back(atoi(bufferVec[1].c_str()));
02295 } else if (lineCnt==3) {
02296 if (bufferVec[0]!="border") error("error in diphon-file:"+diphonName+"\n",-3);
02297 border.push_back(atoi(bufferVec[1].c_str()));
02298 } else if (lineCnt==4) {
02299 if (bufferVec[0]!="t2") error("error in diphon-file:"+diphonName+"\n",-3);
02300 t2.push_back(atoi(bufferVec[1].c_str()));
02301 }
02302 else if (lineCnt>4) {
02303 tmpF0.push_back(atoi(bufferVec[0].c_str()));
02304 }
02305 lineCnt++;
02306 } // if (buffer != "")
02307 } // while (getline(inFile, buffer))
02308 inFile.close();
02309 } else {
02310 error("can't open diphon-file: " + requiredDiphon + "\n", -1);
02311 } // if (inFile.good())
02312 phon++;
02313 } // foreach pho
02314 } // foreach syl
02315
02316 // assign parameters to phones
02317 phon=0;
02318 for (uint syl=0; syl<syllableVec.size(); syl++) {
02319 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
02320 actPhonP = &syllableVec[syl].PhonVec()[pho];
02321 actName = (*actPhonP).Name();
02322 actManner = (*actPhonP).Manner();
02323
02324 if (!(syl==0 && pho==0) && !(syl==syllableVec.size()-1 && pho==syllableVec[syl].PhonNum()-1) ) {
02325 //if (actManner == silence) {
02326 // phonFrameNums.push_back(0);
02327 //} else
02328 if (actManner != stop_voiced && actManner != stop_voiceless) {
02329 // first phon should always be silence
02330 if (phon==0) error("you've found a bug!!!\n",-50);
02331 // cout << actName <<" " << diphonFrameNums[phon-1]-border[phon-1]+border[phon] <<"\n";
02332 phonFrameNum = diphonFrameNums[phon-1] - border[phon-1] + border[phon];
02333 phonBrdFrm = diphonFrameNums[phon-1]-border[phon-1];
02334 (*actPhonP).setSteadyStart(t2[phon-1] - border[phon-1]);
02335 (*actPhonP).setSteadyEnd(diphonFrameNums[phon-1] - border[phon-1] + t1[phon]);
02336 // cout << actName <<" "<<phonFrameNum<<"\n";
02337 } else { // phon is stop
02338 // cout << actName <<" " << diphonFrameNums[phon-1]-border[phon-1]+border[phon]<<"\n";
02339 phonFrameNum = diphonFrameNums[phon-1] - border[phon-1] + border[phon];
02340 phonBrdFrm = diphonFrameNums[phon-1]-border[phon-1];
02341 }
02342
02343 // copy original params
02344 // if (actManner != silence) {
02345
02346 vector<uint> actTmpF0(0);
02347
02348 for (uint phonFrameCnt=0; phonFrameCnt < phonFrameNum; phonFrameCnt++) {
02349 actTmpF0.push_back(tmpF0[allFrameCnt]);
02350 allFrameCnt++;
02351 } // for (uint phonFrameCnt=0; phonFrameCnt<phonFrameNum; phonFrameCnt++)
02352
02353 // check diphonborders for big freq-jumps
02354
02355 // fill param vectors
02356 (*actPhonP).setF0Vec(actTmpF0, glob);
02357
02358 //}
02359 }
02360 phon++;
02361 } // foreach pho
02362 } // foreach syl
02363 }
|
|
|
reads in phones for syllables from inFile. the format of the pho-file is as followed:
|
|
|
assignes only names of phones and syllable borders. It's used for resynthesis if you only want to read in the original intonation contour to print it to a new input-file in connection with
00472 {
00473 string buffer;
00474 // phon tmpPhon;
00475 string name;
00476 const string syllableSeperator = "-";
00477 syllable tmpSyl;
00478
00479 while (getline(inFile, buffer)) {
00480
00481 // a valid line in a pho-file is either a comment
00482 if (buffer != "" && buffer.at(0) != '#') {
00483 vector<string> tokenVec;
00484 tokenVec = string2vec(buffer);
00485
00486
00487 // or a syllable-separator
00488 if (tokenVec[0] == syllableSeperator) { // this is a syllable border
00489 // read the accent-type
00490 if (tokenVec.size()>1) {
00491 tmpSyl.setAccent((ACCENT_TYPE) atoi(tokenVec[1].c_str()));
00492 } else {
00493 error("phrase::readPhoFile: missing accentType in phon-feature near phon "+name+"\n", -3);
00494 }
00495 // push the syllable
00496 syllableVec.push_back(tmpSyl);
00497 // clear the tmp-syllable
00498 tmpSyl.clearPhonVec();
00499
00500 // or a valid pho-file line is a phone-description
00501 } else {
00502 name = tokenVec[0];
00503 phon tmpPhon(name, 0, glob);
00504
00505 tmpSyl.addPhon(tmpPhon);
00506 }
00507
00508 } // if no comment or empty line
00509 } // while (getline(inFile, buffer))
00510 }
|
|
|
reads in amplitudes, formantfreqs and formantbandwidths from diphon-database. It depends on flags whether there are 3 or 5 formantfreqs dynamic. It assignes f0-vals to frames as well as markers for steady state or time of burst.
01995 {
01996
01997 // variable declarations
01998 string actName, nextName, requiredDiphon;
01999 int actFrameNum, actBurstStart, actAspStart;
02000
02001 phon *actPhonP, *nextPhonP;
02002 string diphonType;
02003 string buffer; // tmp-buffer for line in diphon-file
02004 vector<string> bufferVec;
02005 PHON_MANNER actManner;
02006 uint lineCnt=0; // Cnt of lines in diphon-file
02007 vector<uint> diphonFrameNums(0); // list of frameNums foreach diphon
02008
02009 uint phonBrdFrm; // border frame foreach phon
02010 uint phonFrameNum; // frameNum in database of phon
02011 uint allFrameCnt=0; // help Counter for assignment of params to phones
02012
02013 // vectors for markers in diphons
02014 vector<uint> t1(0), t2(0), border(0);
02015 // t1: first transition, t2: 2. transition
02016 // border between diphones
02017
02018 vector<uint> tmpF0;
02019 vector<uint> tmpAmp;
02020 vector<uint> tmpF1, tmpF2, tmpF3, tmpF4, tmpF5;
02021
02022 // borders for shortening of phones
02023 int leftBorder, rightBorder;
02024 uint burstStart, aspStart;
02025
02026 uint phon=0;
02027 string diphonName;
02028 for (uint syl=0; syl<syllableVec.size(); syl++) {
02029 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
02030
02031 actPhonP = &syllableVec[syl].PhonVec()[pho];
02032 if (pho==syllableVec[syl].PhonNum()-1)
02033 if (syl==syllableVec.size()-1)
02034 break;
02035 else
02036 nextPhonP = &syllableVec[syl+1].PhonVec()[0];
02037 else
02038 nextPhonP = &syllableVec[syl].PhonVec()[pho+1];
02039
02040 actName = (*actPhonP).Name();
02041 actFrameNum = (*actPhonP).FrameNum();
02042
02043 nextName = (*nextPhonP).Name();
02044 requiredDiphon = actName + "-" + nextName;
02045 requiredDiphon = diphonBase + "/" + requiredDiphon + ".dip";
02046 ifstream inFile(requiredDiphon.c_str());
02047 if (inFile.good()) {
02048 lineCnt = 0;
02049 while (getline(inFile, buffer)) {
02050 if (buffer != "" && buffer.at(0) != '#') {
02051 bufferVec = string2vec(buffer);
02052 if (lineCnt==0) {
02053 diphonName = bufferVec[1];
02054 }
02055 if (lineCnt==1) {
02056 if (bufferVec[0]!="frameNumber") error("error in diphon-file:"+diphonName+"\n",-3);
02057 diphonFrameNums.push_back(atoi(bufferVec[1].c_str()));
02058 }
02059 else if (lineCnt==2) {
02060 if (bufferVec[0]!="t1") error("error in diphon-file:"+diphonName+"\n",-3);
02061 t1.push_back(atoi(bufferVec[1].c_str()));
02062 } else if (lineCnt==3) {
02063 if (bufferVec[0]!="border") error("error in diphon-file:"+diphonName+"\n",-3);
02064 border.push_back(atoi(bufferVec[1].c_str()));
02065 } else if (lineCnt==4) {
02066 if (bufferVec[0]!="t2") error("error in diphon-file:"+diphonName+"\n",-3);
02067 t2.push_back(atoi(bufferVec[1].c_str()));
02068 }
02069 else if (lineCnt>4) {
02070 tmpF0.push_back(atoi(bufferVec[0].c_str()));
02071 tmpAmp.push_back(atoi(bufferVec[1].c_str()));
02072 tmpF1.push_back(atoi(bufferVec[2].c_str()));
02073 tmpF2.push_back(atoi(bufferVec[3].c_str()));
02074 tmpF3.push_back(atoi(bufferVec[4].c_str()));
02075 if (glob.use4Freq || glob.use5Freq) {
02076 tmpF4.push_back(atoi(bufferVec[5].c_str()));
02077 }
02078 if (glob.use5Freq) {
02079 tmpF5.push_back(atoi(bufferVec[6].c_str()));
02080 }
02081 }
02082 lineCnt++;
02083 } // if (buffer != "")
02084 } // while (getline(inFile, buffer))
02085 inFile.close();
02086 } else {
02087 error("can't open diphon-file: " + requiredDiphon + "\n", -1);
02088 } // if (inFile.good())
02089 phon++;
02090 } // foreach pho
02091 } // foreach syl
02092
02093 // assign parameters to phones
02094 phon=0;
02095 for (uint syl=0; syl<syllableVec.size(); syl++) {
02096 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
02097 actPhonP = &syllableVec[syl].PhonVec()[pho];
02098 actName = (*actPhonP).Name();
02099 actFrameNum = (*actPhonP).FrameNum();
02100
02101 actManner = (*actPhonP).Manner();
02102
02103 if (!(syl==0 && pho==0) && !(syl==syllableVec.size()-1 && pho==syllableVec[syl].PhonNum()-1) ) {
02104 if (actManner != stop_voiced && actManner != stop_voiceless) {
02105 // first phon should always be silence
02106 if (phon==0) error("you've found a bug!!!\n",-50);
02107 phonFrameNum = diphonFrameNums[phon-1] - border[phon-1] + border[phon];
02108 phonBrdFrm = diphonFrameNums[phon-1]-border[phon-1];
02109 (*actPhonP).setSteadyStart(t2[phon-1] - border[phon-1]);
02110 (*actPhonP).setSteadyEnd(diphonFrameNums[phon-1] - border[phon-1] + t1[phon]);
02111
02112 } else { // phon is stop
02113 phonFrameNum = diphonFrameNums[phon-1] - border[phon-1] + border[phon];
02114 phonBrdFrm = diphonFrameNums[phon-1]-border[phon-1];
02115 burstStart = diphonFrameNums[phon-1] - border[phon-1];
02116 aspStart = diphonFrameNums[phon-1] - border[phon-1] + t1[phon];
02117 // cout << actName<<" "<<phonFrameNum<<" "<<burstStart<<" "<<aspStart<<"\n";
02118 }
02119
02120 // copy original params
02121
02122 if (actManner != silence || glob.copysynthesis==true) {
02123 vector<uint> actTmpAmp(0), actTmpF1(0), actTmpF2(0),
02124 actTmpF3(0), actTmpF4(0),
02125 actTmpF5(0);
02126
02127 for (uint phonFrameCnt=0; phonFrameCnt < phonFrameNum; phonFrameCnt++) {
02128 actTmpAmp.push_back(tmpAmp[allFrameCnt]);
02129 actTmpF1.push_back(tmpF1[allFrameCnt]);
02130 actTmpF2.push_back(tmpF2[allFrameCnt]);
02131 actTmpF3.push_back(tmpF3[allFrameCnt]);
02132 if (glob.use4Freq || glob.use5Freq)
02133 actTmpF4.push_back(tmpF4[allFrameCnt]);
02134 if (glob.use5Freq)
02135 actTmpF5.push_back(tmpF5[allFrameCnt]);
02136
02137 allFrameCnt++;
02138 } // for (uint phonFrameCnt=0; phonFrameCnt<phonFrameNum; phonFrameCnt++)
02139 // check diphonborders for big freq-jumps
02140 if (phonBrdFrm!=0 && phonBrdFrm!=phonFrameNum) {
02141 if ((uint)abs((int)(actTmpF1[phonBrdFrm-1]-actTmpF1[phonBrdFrm]))>jumpTolerance && ! glob.warningsOff)
02142 cerr<<"WARNING: readPhonData: big freq jump ("<<abs((int)(actTmpF1[phonBrdFrm-1]-actTmpF1[phonBrdFrm]))<<") for f1 at phon "<<actName<<"\n";
02143 if ((uint)abs((int)(actTmpF2[phonBrdFrm-1]-actTmpF2[phonBrdFrm]))>jumpTolerance && ! glob.warningsOff)
02144 cerr<<"WARNING: readPhonData: big freq jump ("<<abs((int)(actTmpF2[phonBrdFrm-1]-actTmpF2[phonBrdFrm]))<<") for f2 at phon "<<actName<<"\n";
02145 if ((uint)abs((int)(actTmpF3[phonBrdFrm-1]-actTmpF3[phonBrdFrm]))>jumpTolerance && ! glob.warningsOff)
02146 cerr<<"WARNING: readPhonData: big freq jump ("<<abs((int)(actTmpF3[phonBrdFrm-1]-actTmpF3[phonBrdFrm]))<<") for f3 at phon "<<actName<<"\n";
02147 }
02148
02149 // fill param vectors
02150 if (glob.changeDurMethod == STEADY_STATE_CUT || (actManner==stop_voiced || actManner==stop_voiceless)) {
02151 if (actManner != stop_voiced && actManner != stop_voiceless) {
02152 leftBorder = (*actPhonP).SteadyStart();
02153 rightBorder = (*actPhonP).SteadyEnd();
02154 } else { // act phon is stop
02155 leftBorder = 0;
02156 rightBorder = burstStart;
02157 }
02158 // cout << actName<<" "<<(*actPhonP).FrameNum()<<" "<<leftBorder<<" "<<rightBorder<<" "<<phonFrameNum<<"\n";
02159 (*actPhonP).setAmpVec(change_valNum((*actPhonP).Name(), (*actPhonP).FrameNum(), leftBorder, rightBorder, actTmpAmp), glob);
02160 (*actPhonP).setF1Vec(change_valNum((*actPhonP).Name(), (*actPhonP).FrameNum(), leftBorder, rightBorder, actTmpF1), glob);
02161 (*actPhonP).setF2Vec(change_valNum((*actPhonP).Name(), (*actPhonP).FrameNum(), leftBorder, rightBorder, actTmpF2), glob);
02162 (*actPhonP).setF3Vec(change_valNum((*actPhonP).Name(), (*actPhonP).FrameNum(), leftBorder, rightBorder, actTmpF3), glob);
02163 if (glob.use4Freq || glob.use5Freq)
02164 (*actPhonP).setF4Vec(change_valNum((*actPhonP).Name(), (*actPhonP).FrameNum(), leftBorder, rightBorder, actTmpF4), glob);
02165 if (glob.use5Freq)
02166 (*actPhonP).setF5Vec(change_valNum((*actPhonP).Name(), (*actPhonP).FrameNum(), leftBorder, rightBorder, actTmpF5), glob);
02167 } else { // changeDurMethod == TIME_ALIGNED
02168 vector<uint> tmpTimeVec;
02169 for (uint i=0; i<phonFrameNum; i++)
02170 tmpTimeVec.push_back(i*(100/phonFrameNum));
02171 (*actPhonP).setAmpVec(modelTrack((*actPhonP).Name(), (*actPhonP).FrameNum(), actTmpAmp, tmpTimeVec), glob);
02172 (*actPhonP).setF1Vec(modelTrack((*actPhonP).Name(), (*actPhonP).FrameNum(), actTmpF1, tmpTimeVec), glob);
02173 (*actPhonP).setF2Vec(modelTrack((*actPhonP).Name(), (*actPhonP).FrameNum(), actTmpF2, tmpTimeVec), glob);
02174 (*actPhonP).setF3Vec(modelTrack((*actPhonP).Name(), (*actPhonP).FrameNum(), actTmpF3, tmpTimeVec), glob);
02175 if (glob.use4Freq || glob.use5Freq)
02176 (*actPhonP).setF4Vec(modelTrack((*actPhonP).Name(), (*actPhonP).FrameNum(), actTmpF4, tmpTimeVec), glob);
02177 if (glob.use5Freq)
02178 (*actPhonP).setF5Vec(modelTrack((*actPhonP).Name(), (*actPhonP).FrameNum(), actTmpF5, tmpTimeVec), glob);
02179
02180 }
02181 if (actManner == stop_voiced || actManner == stop_voiceless) {
02182 // cout <<actName<<": bS:"<<burstStart<<", aS:"<<aspStart<<"\n";
02183 if (glob.copysynthesis) {
02184 if (burstStart < (phonFrameNum-actFrameNum)) {
02185 actBurstStart = 0;
02186 actAspStart = aspStart - burstStart;
02187 } else {
02188 actBurstStart = burstStart - (phonFrameNum-actFrameNum);
02189 actAspStart = aspStart - (phonFrameNum-actFrameNum);
02190 }
02191 } else {
02192 if (actFrameNum>=2) {
02193 actBurstStart = actFrameNum-2;
02194 actAspStart = actFrameNum-1;
02195 } else if (actFrameNum==1) {
02196 actBurstStart = 0;
02197 actAspStart = 1;
02198 } else {
02199 actBurstStart = 0;
02200 actAspStart = 0;
02201 }
02202 }
02203 (*actPhonP).setBurstStart(actBurstStart);
02204 (*actPhonP).setAspStart(actAspStart);
02205 // cout <<actName<<", bS:"<<actBurstStart<<", aS: "<<actAspStart<<", frameLength: "<<actFrameNum<<", dataBaseLength: "<<phonFrameNum<<"\n";
02206
02207 }
02208
02209 } else {
02210 // for silent parts interpolate the formant tracks to avoid freq-jumps
02211 uint startF1=tmpF1[allFrameCnt-1]; uint endF1=tmpF1[allFrameCnt];
02212 (*actPhonP).setF1Vec(interpolateLinear((*actPhonP).FrameNum(), startF1, endF1), glob);
02213 uint startF2=tmpF2[allFrameCnt-1]; uint endF2=tmpF2[allFrameCnt];
02214 (*actPhonP).setF2Vec(interpolateLinear((*actPhonP).FrameNum(), startF2, endF2), glob);
02215 uint startF3=tmpF3[allFrameCnt-1]; uint endF3=tmpF3[allFrameCnt];
02216 (*actPhonP).setF3Vec(interpolateLinear((*actPhonP).FrameNum(), startF3, endF3), glob);
02217 } // if (actManner != silence)
02218 }
02219 if (syl==syllableVec.size()-1 && pho==syllableVec[syl].PhonNum()-1) {
02220 // for silence at end continue formant-tracks to avoid freq-jumps
02221 // cout <<tmpF1[allFrameCnt-1]<<"\n";
02222 uint startF1=tmpF1[allFrameCnt-1];
02223 (*actPhonP).setF1Vec(interpolateLinear((*actPhonP).FrameNum(), startF1, startF1), glob);
02224 uint startF2=tmpF2[allFrameCnt-1];
02225 (*actPhonP).setF2Vec(interpolateLinear((*actPhonP).FrameNum(), startF2, startF2), glob);
02226 uint startF3=tmpF3[allFrameCnt-1];
02227 (*actPhonP).setF3Vec(interpolateLinear((*actPhonP).FrameNum(), startF3, startF3), glob);
02228 }
02229 phon++;
02230 } // foreach pho
02231 } // foreach syl
02232 }
|
|
|
adds rounded Lips characteristices. All formants get lowered acording to rate.
01380 {
01381 phon *pP;
01382 for (uint syl=0; syl<syllableVec.size(); syl++) {
01383 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01384 pP = &syllableVec[syl].PhonVec()[pho];
01385 if ((*pP).Type() == vowel) {
01386 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01387 tmpPhonFeatures.rounded = true;
01388 tmpPhonFeatures.roundedRate = rate;
01389 (*pP).setPhonFeatures(tmpPhonFeatures);
01390 }
01391 }
01392 }
01393 if (glob.verbose)
01394 cerr <<"added lip-rounding at rate: "<<rate<<"\n";
01395 }
|
|
|
smoothing function: all f0Vals get centerFiltered around winsize.
00781 {
00782 for (uint syl=0; syl<syllableVec.size(); syl++) {
00783 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
00784 if (syllableVec[syl].PhonVec()[pho].Voiced() && syllableVec[syl].PhonVec()[pho].Manner() != stop_voiced) {
00785 syllableVec[syl].PhonVec()[pho].smootheF0(winsize);
00786 }
00787 }
00788 }
00789 if (glob.verbose)
00790 cerr << "smoothing f0 with winSize: "<<winsize<<"\n";
00791 }
|
|
|
adds spread Lips characteristices. The first 2 formants get shifted upwards according to rate.
01363 {
01364 phon *pP;
01365 for (uint syl=0; syl<syllableVec.size(); syl++) {
01366 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01367 pP = &syllableVec[syl].PhonVec()[pho];
01368 if ((*pP).Type()==vowel) {
01369 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01370 tmpPhonFeatures.spread = true;
01371 tmpPhonFeatures.spreadRate = rate;
01372 (*pP).setPhonFeatures(tmpPhonFeatures);
01373 }
01374 }
01375 }
01376 if (glob.verbose)
01377 cerr <<"added lip-spreading at rate: "<<rate<<"\n";
01378 }
|
|
|
blur the vowel precision. all formantVals in f1 and f2 of vowels in syllables with get displaced towards the center freqencies by subtracting the percentage by <rate> and distance to center frequency.
01345 {
01346 phon *pP;
01347 for (uint syl=0; syl<syllableVec.size(); syl++) {
01348 if (syllableVec[syl].Accent()==accent) {
01349 for (uint pho=0; pho<syllableVec[syl].PhonNum(); pho++) {
01350 pP = &syllableVec[syl].PhonVec()[pho];
01351 if ((*pP).Type() == vowel) {
01352 PHON_FEATURES tmpPhonFeatures= (*pP).PhonFeatures();
01353 tmpPhonFeatures.undershoot = true;
01354 tmpPhonFeatures.undershootRate = rate;
01355 (*pP).setPhonFeatures(tmpPhonFeatures);
01356 }
01357 }
01358 }
01359 }
01360 if (glob.verbose)
01361 cerr <<"added undershoot at rate: "<<rate<<" for accent-type: "<<(uint)accent<<"\n";
01362 }
|
|
|
|
|
|
|
|
|
|