encoder.cpp

mpeg4实现的软件代码

    data_write(&prosodyEnable,1,1,YY,fout);
    data_write(&videoEnable,1,1,YY,fout);
    data_write(&lipShapeEnable,1,1,YY,fout);
    data_write(&trickModeEnable,1,1,YY,fout);
    if(YY[1]!=8) { c=0; i=YY[1]; data_write(&c,i,1,YY,fout); }
}

void ttsSequence::clear_ttsSentence()
{
    ttsSentence *s=sentence;
    ttsProsody *p=s->prosody;

    if(!s->silence) {
		free(s->ttsText); 
		s->ttsText=NULL;
		if(prosodyEnable) {
			free(p->phonemeSymbols); p->phonemeSymbols=NULL;
			if(get_durEn())
			{ free(p->durEachPhoneme); p->durEachPhoneme=NULL; }
			if(get_f0En())
			{ free(p->f0ContourEachPhoneme); p->f0ContourEachPhoneme=NULL; }
			if(get_enEn()) {
				free(p->energyContourEachPhoneme); 
				p->energyContourEachPhoneme=NULL;
			}
	    }
    	if(lipShapeEnable) {
	    free(s->LipShapeInSentence); s->LipShapeInSentence=NULL;
	    free(s->lipShape); s->lipShape=NULL;
		}   
	}
}


void ttsSequence::save_ttsSentence(FILE *fout)
{
    ttsSentence *s=sentence;
    ttsProsody *p=s->prosody;
    int i, j, Ts, Tn, N, nphone; char c;

    if(videoEnable) {
		Ts=(int)(s->startTime*1000);  Tn=iframe*itime;
		N=(Ts-Tn)/itime; 
		for(i=0; i<N; i++) {
			data_write(&s->tts_sentence_start_code,32,1,YY,fout);
			data_write(&s->ttsSentenceID,10,1,YY,fout);
			c=1; data_write(&c,1,1,YY,fout);
			data_write(&itime,12,1,YY,fout);
			if(YY[1]!=8) { c=0; j=YY[1]; data_write(&c,j,1,YY,fout); }
	    } iframe+=N;
		Tn=s->offset+s->sentenceDuration; N=Tn/itime;
		if((Tn%itime)!=0) N++;
	} 
    else N=1;

    for(i=0; i<N; i++) {	
    	data_write(&s->tts_sentence_start_code,32,1,YY,fout);
    	data_write(&s->ttsSentenceID,10,1,YY,fout);
    	data_write(&s->silence,1,1,YY,fout);
    	if(s->silence) data_write(&s->silenceDuration,12,1,YY,fout);
    	else {
			if(genderEnable) data_write(&s->gender,1,1,YY,fout);
			if(ageEnable) data_write(&s->age,3,1,YY,fout);
			if(!videoEnable && speechRateEnable)
				data_write(&s->speechRate,4,1,YY,fout);
    	    data_write(&s->lengthText,12,1,YY,fout);
    	    data_write(s->ttsText,8,s->lengthText,YY,fout);
			if(prosodyEnable) {
				data_write(&p->durEnable,1,1,YY,fout);
				data_write(&p->f0ContourEnable,1,1,YY,fout);
				data_write(&p->energyContourEnable,1,1,YY,fout);
				nphone=p->numberPhonemes;
				data_write(&nphone,10,1,YY,fout);
				data_write(&p->phonemeSymbolsLength,13,1,YY,fout);
				data_write(p->phonemeSymbols,8,p->phonemeSymbolsLength,YY,fout);
				if(get_durEn())
	    			data_write(p->durEachPhoneme,12,nphone,YY,fout);
				if(get_f0En())
	    			data_write(p->f0ContourEachPhoneme,8,nphone*3,YY,fout);
				if(get_enEn())
	    			data_write(p->energyContourEachPhoneme,8,nphone*3,YY,fout);
	        }
    	    if(videoEnable) {
				data_write(&s->sentenceDuration,16,1,YY,fout);
				data_write(&s->positionInSentence,16,1,YY,fout);
				data_write(&s->offset,10,1,YY,fout);
				if(s->offset) s->positionInSentence+=(itime-s->offset);
				else s->positionInSentence+=itime;
				s->offset=0;
	        }
    	    if(lipShapeEnable) {
				j=s->numberLipShape;
				data_write(&j,10,1,YY,fout);
				data_write(s->LipShapeInSentence,16,j,YY,fout);
				data_write(s->lipShape,8,j,YY,fout);
		    }   
		}
        if(YY[1]!=8) { c=0; j=YY[1]; data_write(&c,j,1,YY,fout); }
	} 
    if(videoEnable) iframe+=N;
}

void ttsSequence::data_write(char *A, int nbits, int n, char *B, FILE *fout)
{
    int i, nx, ny;
    unsigned char X,Y;
    short W[9]={ 0,1,3,7,0x0F,0x01F,0x03F,0x07F,0x0FF };

    Y=B[0]; ny=B[1];
    for(i=0; i<n; i++) {
        nx=nbits;
        while(nx!=0) {
            if(ny>nx) { X=*(A+i) & W[nx]; Y=Y | (X << (ny-nx)); ny-=nx; nx=0; }
            else {
                X=(*(A+i)) >> (nx-ny); Y=Y | (X & W[ny]); nx-=ny; ny=8;
                fwrite(&Y,1,1,fout); Y=0;
            }   }
        } B[0]=Y; B[1]=ny;
}

void ttsSequence::data_write(short *A, int nbits, int n, char *B, FILE *fout)
{
    int i, nx, ny;
    unsigned short X; unsigned char Y;
    short W[9]={ 0,1,3,7,0x0F,0x01F,0x03F,0x07F,0x0FF };

    Y=B[0]; ny=B[1];
    for(i=0; i<n; i++) {
        nx=nbits;
        while(nx!=0) {
            if(ny>nx) { X=*(A+i) & W[nx]; Y=Y | (X << (ny-nx)); ny-=nx; nx=0; }
            else {
                X=(*(A+i)) >> (nx-ny); Y=Y | (X & W[ny]); nx-=ny; ny=8;
                fwrite(&Y,1,1,fout); Y=0;
            }   }
        } B[0]=Y; B[1]=ny;
}

void ttsSequence::data_write(int *A, int nbits, int n, char *B, FILE *fout)
{
    int i, nx, ny;
    unsigned int X; unsigned char Y;
    short W[9]={ 0,1,3,7,0x0F,0x01F,0x03F,0x07F,0x0FF };

    Y=B[0]; ny=B[1];
    for(i=0; i<n; i++) {
        nx=nbits;
        while(nx!=0) {
            if(ny>nx) { X=*(A+i) & W[nx]; Y=Y | (X << (ny-nx)); ny-=nx; nx=0; }
            else {
                X=(*(A+i)) >> (nx-ny); Y=Y | (X & W[ny]); nx-=ny; ny=8;
                fwrite(&Y,1,1,fout); Y=0;
            }
		}
    } B[0]=Y; B[1]=ny;
}

int rd_txt(FILE *fp, char *Text);
int read_prosody(FILE *fp,short Sphon[][6],short Dur[],short F0[][3],short En[][3]);
void read_video(FILE *fp,float *stime,int *dur);

int Niframe, v_st, v_ed, sp_st, sp_ed;
float Itime;

void Encoder(int E_Lang_v, int E_Gender, int E_Gender_v, int E_Age, int E_Age_v, int E_Spch, int E_Spch_v, int E_Proso, int E_Proso_d, int E_Proso_F0, int E_Proso_e, int E_Video, int E_Lip, int E_Trick )
{
    int i, itime, Nphone, dur;  
    short Dur[800], F0[800][3], En[800][3];  
    float stime;
    char fname[30],Text[800];
	short Sphone[800][6];
    FILE *ftext, *fvideo, *fprosody, *fout; 

	G_Lang_v = E_Lang_v; G_Gender = E_Gender; G_Gender_v = E_Gender_v;
	G_Age = E_Age; G_Age_v = E_Age_v;  G_Spch = E_Spch; G_Spch_v=E_Spch_v; 
	G_Proso =  E_Proso;  G_Proso_d = E_Proso_d; G_Proso_F0 = E_Proso_F0; G_Proso_e = E_Proso_e; 
	G_Video = E_Video; G_Lip = E_Lip; G_Trick = E_Trick;

	/* We assume time duration between adjacent I-frames is 0.5sec */
    itime=500;  /* msec */

    i=0;
    ttsSequence *ttsSeq = new ttsSequence(1);

    ttsSeq->init();

    ttsSentence *ttsSnt = new ttsSentence(++i);
    ttsSeq->AddSentence(ttsSnt);
    ttsSeq->Sntinit();

    YY[0]=0; YY[1]=8;

	/* Check Data files */

    if((ftext=fopen("ko_text.dat","r"))==NULL)
	{ ::MessageBox(NULL, "Error: ko_text.dat open", "ERROR", MB_OK); exit(1); }

    if(ttsSeq->get_prosodyEn()) {
		if(ttsSeq->get_gender()) strcpy(fname,"ko_m-prosody.dat");
		else strcpy(fname,"ko_f-prosody.dat");
    	if((fprosody=fopen(fname,"r"))==NULL)
		{ ::MessageBox(NULL, "Error: prosody data open", "ERROR", MB_OK); exit(1); }
	}
    if(ttsSeq->get_videoEn()) {
		if(ttsSeq->get_gender()) strcpy(fname,"ko_m-video.dat");
		else strcpy(fname,"ko_f-video.dat");
    	if((fvideo=fopen(fname,"r"))==NULL)
		{ ::MessageBox(NULL, "Err: video data open", "ERROR", MB_OK); exit(1); }
	}

    fout=fopen("mpeg_tts.dat","w+b");
    ttsSeq->save_ttsSequence(fout);
    while( (i=rd_txt(ftext,Text)) > 1 ) {
		ttsSeq->put_text(Text);
		i=ttsSeq->get_text(Text); //printf("%d %s\n",i,Text);

    	if(ttsSeq->get_prosodyEn()) {
			Nphone=read_prosody(fprosody,Sphone,Dur,F0,En);
			if( Dur[0] == -1000 )	ttsSeq->set_durEn(0);
			if( F0[0][0] == -1000 )	ttsSeq->set_f0En(0);
			if( En[0][0] == -1000 )	ttsSeq->set_enEn(0);
			ttsSeq->put_prosody(Nphone,Sphone,Dur,F0,En);
	    }

    	if(ttsSeq->get_videoEn()) {
			read_video(fvideo,&stime,&dur);
			ttsSeq->put_position(itime,stime,dur);
	    }

    	ttsSeq->save_ttsSentence(fout);
		ttsSeq->clear_ttsSentence();
	} 
	fclose(ftext); fclose(fout);

}

int rd_txt(FILE *fp, char *Text)
{
    int sentnce_end=0, i=0, ch; char pchr;

    pchr=32;
    while( (ch=fgetc(fp)) != EOF && i< 800 ) {
        if(ch=='\n' || ch=='\r' || ch == '\0' || ch =='\t') ch=32;
        if(ch=='"' || ch=='\'' || ch=='{' || ch=='}' || ch=='[' || ch==']' ||
           ch=='<' || ch=='>' || ch=='(' || ch==')' || ch==',' ) continue;
        if(ch != 32 || pchr != 32) Text[i++]=ch;
        if(ch=='.') {
            ch=fgetc(fp); if(!isdigit(ch)) { sentnce_end=1; break; }
            Text[i++]=ch;
            }
        pchr=ch; if( ch=='?' || ch=='!') { sentnce_end=1; break; }
        }
    if(sentnce_end == 0) { if(Text[i-1]==32) i--;  Text[i++]='.'; }
    Text[i]=0; return(i);
}

int read_prosody(FILE *fp,short Sphon[][6],short Dur[],short F0[][3],short En[][3])
{
	int i, j, k, m[10], n=0, np, nIPA;
    char line[100], str[15][20], *tokn;
    char *field[]={ "IPA_phones","dur","F0_1","F0_2","F0_3","En_1","En_2","En_3" };

    for(i=0; i<8; i++) m[i]=-1;
    fgets(line,90,fp);
    for(i=0, tokn=strtok(line," \n"); tokn!=NULL; tokn=strtok(NULL," \n")) 
	strcpy(str[i++],tokn);
    np=i-1;
    if(strcmp(str[0],"begin")!=0) { ::MessageBox(NULL, "prosody file err", "Error", MB_OK); exit(1); }
    for(j=1; j<i; j++) {
	for(k=0; k<8; k++) 
	    if(strcmp(str[j],field[k])==0) { m[k]=j-1; break; }
	}

    while(fgets(line,90,fp)!=NULL) {
    	for(i=0, tokn=strtok(line," \n"); tokn!=NULL; tokn=strtok(NULL," \n")) 
	    strcpy(str[i++],tokn);
		if(strcmp(str[0],"end")==0) break;
		nIPA=i-(np-1);
		for(j=0; j<nIPA; j++) {
			sscanf(str[m[0]+j],"%x",&k); Sphon[n][j]=k;
			}
		Sphon[n][j]=0;
		if(m[1]!=-1) Dur[n]=(int)(atof(str[m[1]+nIPA-1])*1000); 
		else Dur[0] = -1000;
		for(i=2; i<5; i++) {
			if(m[i]!=-1) F0[n][i-2]=atoi(str[m[i]+nIPA-1])/2; 
			else { F0[0][0] = -1000; break; }
		}
		for(i=5; i<8; i++) {
			if(m[i]!=-1) En[n][i-5]=atoi(str[m[i]+nIPA-1]); 
			else { En[0][0] = -1000; break;	}
		}
		n++;
	}
    return(n);

}

void read_video(FILE *fp,float *stime,int *dur)
{
    int i; float t;
    char line[100], str[10][10], *tokn;
    
    fgets(line,90,fp);
    for(i=0, tokn=strtok(line," \n"); tokn!=NULL; tokn=strtok(NULL," \n")) 
	strcpy(str[i++],tokn);
    if(strcmp(str[0],"begin")!=0) { ::MessageBox(NULL, "video file err", "Error", MB_OK); exit(1); }
    fgets(line,90,fp);  tokn=strtok(line," \n"); *stime=(float)atof(tokn);
    fgets(line,90,fp);  tokn=strtok(line," \n"); t=(float)atof(tokn);
    *dur=(int)((t-*stime)*1000);
    fgets(line,90,fp);  tokn=strtok(line," \n");
    if(strcmp(tokn,"end")!=0) { ::MessageBox(NULL, "video file err", "Error", MB_OK); exit(1); }
}