VorbisPsy *vorbis_psy_init(int rate, int n)
{
long i,j,lo=-99,hi=1;
VorbisPsy *p = speex_alloc(sizeof(VorbisPsy));
memset(p,0,sizeof(*p));
p->n = n;
spx_drft_init(&p->lookup, n);
p->bark = speex_alloc(n*sizeof(*p->bark));
p->rate=rate;
p->vi = &example_tuning;
/* BH4 window */
p->window = speex_alloc(sizeof(*p->window)*n);
float a0 = .35875f;
float a1 = .48829f;
float a2 = .14128f;
float a3 = .01168f;
for(i=0;i<n;i++)
p->window[i] = //a0 - a1*cos(2.*M_PI/n*(i+.5)) + a2*cos(4.*M_PI/n*(i+.5)) - a3*cos(6.*M_PI/n*(i+.5));
sin((i+.5)/n * M_PI)*sin((i+.5)/n * M_PI);
/* bark scale lookups */
for(i=0;i<n;i++){
float bark=toBARK(rate/(2*n)*i);
for(;lo+p->vi->noisewindowlomin<i &&
toBARK(rate/(2*n)*lo)<(bark-p->vi->noisewindowlo);lo++);
for(;hi<=n && (hi<i+p->vi->noisewindowhimin ||
toBARK(rate/(2*n)*hi)<(bark+p->vi->noisewindowhi));hi++);
p->bark[i]=((lo-1)<<16)+(hi-1);
}
/* set up rolling noise median */
p->noiseoffset=speex_alloc(n*sizeof(*p->noiseoffset));
for(i=0;i<n;i++){
float halfoc=toOC((i+.5)*rate/(2.*n))*2.;
int inthalfoc;
float del;
if(halfoc<0)halfoc=0;
if(halfoc>=P_BANDS-1)halfoc=P_BANDS-1;
inthalfoc=(int)halfoc;
del=halfoc-inthalfoc;
p->noiseoffset[i]=
p->vi->noiseoff[inthalfoc]*(1.-del) +
p->vi->noiseoff[inthalfoc+1]*del;
}
#if 0
_analysis_output_always("noiseoff0",ls,p->noiseoffset,n,1,0,0);
#endif
return p;
}
vorbis_look_psy_global *_vp_global_look(vorbis_info *vi){
int i,j;
codec_setup_info *ci=vi->codec_setup;
vorbis_info_psy_global *gi=ci->psy_g_param;
vorbis_look_psy_global *look=_ogg_calloc(1,sizeof(vorbis_look_psy_global));
int shiftoc=rint(log(gi->eighth_octave_lines*8)/log(2))-1;
look->decaylines=toOC(96000.f)*(1<<(shiftoc+1))+.5f; /* max sample
rate of
192000kHz
for now */
look->decay=_ogg_calloc(vi->channels,sizeof(float *));
for(i=0;i<vi->channels;i++){
look->decay[i]=_ogg_calloc(look->decaylines,sizeof(float));
for(j=0;j<look->decaylines;j++)
look->decay[i][j]=-9999.;
}
look->channels=vi->channels;
look->ampmax=-9999.;
look->gi=gi;
return(look);
}
void _vp_psy_init(vorbis_look_psy *p,vorbis_info_psy *vi,
vorbis_info_psy_global *gi,int n,long rate){
long i,j,lo=-99,hi=1;
long maxoc;
memset(p,0,sizeof(*p));
p->eighth_octave_lines=gi->eighth_octave_lines;
p->shiftoc=rint(log(gi->eighth_octave_lines*8.f)/log(2.f))-1;
p->firstoc=toOC(.25f*rate*.5/n)*(1<<(p->shiftoc+1))-gi->eighth_octave_lines;
maxoc=toOC((n+.25f)*rate*.5/n)*(1<<(p->shiftoc+1))+.5f;
p->total_octave_lines=maxoc-p->firstoc+1;
p->ath=(float*)_ogg_malloc(n*sizeof(*p->ath));
p->octave=(long*)_ogg_malloc(n*sizeof(*p->octave));
p->bark=(long*)_ogg_malloc(n*sizeof(*p->bark));
p->vi=vi;
p->n=n;
p->rate=rate;
/* AoTuV HF weighting */
p->m_val = 1.;
if(rate < 26000) p->m_val = 0;
else if(rate < 38000) p->m_val = .94; /* 32kHz */
else if(rate > 46000) p->m_val = 1.275; /* 48kHz */
/* set up the lookups for a given blocksize and sample rate */
for(i=0,j=0;i<MAX_ATH-1;i++){
int endpos=rint(fromOC((i+1)*.125-2.)*2*n/rate);
float base=ATH[i];
if(j<endpos){
float delta=(ATH[i+1]-base)/(endpos-j);
for(;j<endpos && j<n;j++){
p->ath[j]=base+100.;
base+=delta;
}
}
}
for(i=0;i<n;i++){
float bark=toBARK(rate/(2*n)*i);
for(;lo+vi->noisewindowlomin<i &&
toBARK(rate/(2*n)*lo)<(bark-vi->noisewindowlo);lo++);
for(;hi<=n && (hi<i+vi->noisewindowhimin ||
toBARK(rate/(2*n)*hi)<(bark+vi->noisewindowhi));hi++);
p->bark[i]=((lo-1)<<16)+(hi-1);
}
for(i=0;i<n;i++)
p->octave[i]=toOC((i+.25f)*.5*rate/n)*(1<<(p->shiftoc+1))+.5f;
p->tonecurves=setup_tone_curves(vi->toneatt,rate*.5/n,n,
vi->tone_centerboost,vi->tone_decay);
/* set up rolling noise median */
p->noiseoffset=(float**)_ogg_malloc(P_NOISECURVES*sizeof(*p->noiseoffset));
for(i=0;i<P_NOISECURVES;i++)
p->noiseoffset[i]=(float*)_ogg_malloc(n*sizeof(**p->noiseoffset));
for(i=0;i<n;i++){
float halfoc=toOC((i+.5)*rate/(2.*n))*2.;
int inthalfoc;
float del;
if(halfoc<0)halfoc=0;
if(halfoc>=P_BANDS-1)halfoc=P_BANDS-1;
inthalfoc=(int)halfoc;
del=halfoc-inthalfoc;
for(j=0;j<P_NOISECURVES;j++)
p->noiseoffset[j][i]=
p->vi->noiseoff[j][inthalfoc]*(1.-del) +
p->vi->noiseoff[j][inthalfoc+1]*del;
}
#if 0
{
static int ls=0;
_analysis_output_always("noiseoff0",ls,p->noiseoffset[0],n,1,0,0);
_analysis_output_always("noiseoff1",ls,p->noiseoffset[1],n,1,0,0);
_analysis_output_always("noiseoff2",ls++,p->noiseoffset[2],n,1,0,0);
}
#endif
}
void _vp_psy_init(vorbis_look_psy *p,vorbis_info_psy *vi,
vorbis_info_psy_global *gi,int n,long rate){
long i,j,k,lo=0,hi=0;
long maxoc;
memset(p,0,sizeof(vorbis_look_psy));
p->eighth_octave_lines=gi->eighth_octave_lines;
p->shiftoc=rint(log(gi->eighth_octave_lines*8)/log(2))-1;
p->firstoc=toOC(.25f*rate/n)*(1<<(p->shiftoc+1))-gi->eighth_octave_lines;
maxoc=toOC((n*.5f-.25f)*rate/n)*(1<<(p->shiftoc+1))+.5f;
p->total_octave_lines=maxoc-p->firstoc+1;
if(vi->ath)
p->ath=_ogg_malloc(n*sizeof(float));
p->octave=_ogg_malloc(n*sizeof(long));
p->bark=_ogg_malloc(n*sizeof(unsigned long));
p->vi=vi;
p->n=n;
p->rate=rate;
/* set up the lookups for a given blocksize and sample rate */
if(vi->ath)
set_curve(vi->ath, p->ath,n,rate);
for(i=0;i<n;i++){
float bark=toBARK(rate/(2*n)*i);
for(;lo+vi->noisewindowlomin<i &&
toBARK(rate/(2*n)*lo)<(bark-vi->noisewindowlo);lo++);
for(;hi<n && (hi<i+vi->noisewindowhimin ||
toBARK(rate/(2*n)*hi)<(bark+vi->noisewindowhi));hi++);
p->bark[i]=(hi<<16)+lo;
}
for(i=0;i<n;i++)
p->octave[i]=toOC((i*.5f+.25f)*rate/n)*(1<<(p->shiftoc+1))+.5f;
p->tonecurves=_ogg_malloc(P_BANDS*sizeof(float **));
p->noisethresh=_ogg_malloc(n*sizeof(float));
p->noiseoffset=_ogg_malloc(n*sizeof(float));
for(i=0;i<P_BANDS;i++)
p->tonecurves[i]=_ogg_malloc(P_LEVELS*sizeof(float *));
for(i=0;i<P_BANDS;i++)
for(j=0;j<P_LEVELS;j++)
p->tonecurves[i][j]=_ogg_malloc((EHMER_MAX+2)*sizeof(float));
/* OK, yeah, this was a silly way to do it */
memcpy(p->tonecurves[0][4]+2,tone_125_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[0][6]+2,tone_125_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[0][8]+2,tone_125_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[0][10]+2,tone_125_100dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[2][4]+2,tone_125_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[2][6]+2,tone_125_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[2][8]+2,tone_125_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[2][10]+2,tone_125_100dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[4][4]+2,tone_250_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[4][6]+2,tone_250_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[4][8]+2,tone_250_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[4][10]+2,tone_250_100dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[6][4]+2,tone_500_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[6][6]+2,tone_500_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[6][8]+2,tone_500_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[6][10]+2,tone_500_100dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[8][4]+2,tone_1000_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[8][6]+2,tone_1000_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[8][8]+2,tone_1000_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[8][10]+2,tone_1000_100dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[10][4]+2,tone_2000_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[10][6]+2,tone_2000_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[10][8]+2,tone_2000_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[10][10]+2,tone_2000_100dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[12][4]+2,tone_4000_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[12][6]+2,tone_4000_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[12][8]+2,tone_4000_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[12][10]+2,tone_4000_100dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[14][4]+2,tone_8000_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[14][6]+2,tone_8000_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[14][8]+2,tone_8000_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[14][10]+2,tone_8000_100dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[16][4]+2,tone_8000_40dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[16][6]+2,tone_8000_60dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[16][8]+2,tone_8000_80dB_SL,sizeof(float)*EHMER_MAX);
memcpy(p->tonecurves[16][10]+2,tone_8000_100dB_SL,sizeof(float)*EHMER_MAX);
/* value limit the tonal masking curves; the peakatt not only
optionally specifies maximum dynamic depth, but also [always]
limits the masking curves to a minimum depth */
for(i=0;i<P_BANDS;i+=2)
for(j=4;j<P_LEVELS;j+=2)
for(k=2;k<EHMER_MAX+2;k++)
p->tonecurves[i][j][k]+=vi->tone_masteratt;
/* interpolate curves between */
for(i=1;i<P_BANDS;i+=2)
for(j=4;j<P_LEVELS;j+=2){
memcpy(p->tonecurves[i][j]+2,p->tonecurves[i-1][j]+2,EHMER_MAX*sizeof(float));
/*interp_curve(p->tonecurves[i][j],
p->tonecurves[i-1][j],
p->tonecurves[i+1][j],.5);*/
min_curve(p->tonecurves[i][j]+2,p->tonecurves[i+1][j]+2);
}
/* set up the final curves */
for(i=0;i<P_BANDS;i++)
setup_curve(p->tonecurves[i],i,vi->toneatt->block[i]);
if(vi->curvelimitp){
/* value limit the tonal masking curves; the peakatt not only
optionally specifies maximum dynamic depth, but also [always]
limits the masking curves to a minimum depth */
for(i=0;i<P_BANDS;i++)
for(j=0;j<P_LEVELS;j++){
for(k=2;k<EHMER_OFFSET+2+vi->curvelimitp;k++)
if(p->tonecurves[i][j][k]> vi->peakatt->block[i][j])
p->tonecurves[i][j][k]= vi->peakatt->block[i][j];
else
break;
}
}
if(vi->peakattp) /* we limit depth only optionally */
for(i=0;i<P_BANDS;i++)
for(j=0;j<P_LEVELS;j++)
if(p->tonecurves[i][j][EHMER_OFFSET+2]< vi->peakatt->block[i][j])
p->tonecurves[i][j][EHMER_OFFSET+2]= vi->peakatt->block[i][j];
/* but guarding is mandatory */
for(i=0;i<P_BANDS;i++)
for(j=0;j<P_LEVELS;j++)
if(p->tonecurves[i][j][EHMER_OFFSET+2]< vi->tone_maxatt)
p->tonecurves[i][j][EHMER_OFFSET+2]= vi->tone_maxatt;
/* set up rolling noise median */
for(i=0;i<n;i++){
float halfoc=toOC((i+.5)*rate/(2.*n))*2.;
int inthalfoc;
float del;
if(halfoc<0)halfoc=0;
if(halfoc>=P_BANDS-1)halfoc=P_BANDS-1;
inthalfoc=(int)halfoc;
del=halfoc-inthalfoc;
p->noisethresh[i]=((p->vi->noisethresh[inthalfoc]*(1.-del) +
p->vi->noisethresh[inthalfoc+1]*del))*2.f-1.f;
p->noiseoffset[i]=
p->vi->noiseoff[inthalfoc]*(1.-del) +
p->vi->noiseoff[inthalfoc+1]*del;
}
analysis_noisy=1;
_analysis_output("noiseoff",0,p->noiseoffset,n,1,0);
_analysis_output("noisethresh",0,p->noisethresh,n,1,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_63Hz",i,p->tonecurves[0][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_88Hz",i,p->tonecurves[1][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_125Hz",i,p->tonecurves[2][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_170Hz",i,p->tonecurves[3][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_250Hz",i,p->tonecurves[4][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_350Hz",i,p->tonecurves[5][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_500Hz",i,p->tonecurves[6][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_700Hz",i,p->tonecurves[7][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_1kHz",i,p->tonecurves[8][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_1.4Hz",i,p->tonecurves[9][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_2kHz",i,p->tonecurves[10][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_2.4kHz",i,p->tonecurves[11][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_4kHz",i,p->tonecurves[12][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_5.6kHz",i,p->tonecurves[13][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_8kHz",i,p->tonecurves[14][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_11.5kHz",i,p->tonecurves[15][i]+2,EHMER_MAX,0,0);
for(i=0;i<P_LEVELS;i++)
_analysis_output("curve_16kHz",i,p->tonecurves[16][i]+2,EHMER_MAX,0,0);
analysis_noisy=1;
}
