1 /********************************************************************
3 * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE. *
4 * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS *
5 * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
6 * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. *
8 * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009 *
9 * by the Xiph.Org Foundation http://www.xiph.org/ *
11 ********************************************************************
13 function: psychoacoustics not including preecho
16 ********************************************************************/
21 #include "vorbis/codec.h"
22 #include "codec_internal.h"
32 #define NEGINF -9999.f
33 static const double stereo_threshholds[]={0.0, .5, 1.0, 1.5, 2.5, 4.5, 8.5, 16.5, 9e10};
34 static const double stereo_threshholds_limited[]={0.0, .5, 1.0, 1.5, 2.0, 2.5, 4.5, 8.5, 9e10};
36 vorbis_look_psy_global *_vp_global_look(vorbis_info *vi){
37 codec_setup_info *ci=vi->codec_setup;
38 vorbis_info_psy_global *gi=&ci->psy_g_param;
39 vorbis_look_psy_global *look=_ogg_calloc(1,sizeof(*look));
41 look->channels=vi->channels;
48 void _vp_global_free(vorbis_look_psy_global *look){
50 memset(look,0,sizeof(*look));
55 void _vi_gpsy_free(vorbis_info_psy_global *i){
57 memset(i,0,sizeof(*i));
62 void _vi_psy_free(vorbis_info_psy *i){
64 memset(i,0,sizeof(*i));
69 static void min_curve(float *c,
72 for(i=0;i<EHMER_MAX;i++)if(c2[i]<c[i])c[i]=c2[i];
74 static void max_curve(float *c,
77 for(i=0;i<EHMER_MAX;i++)if(c2[i]>c[i])c[i]=c2[i];
80 static void attenuate_curve(float *c,float att){
82 for(i=0;i<EHMER_MAX;i++)
86 static float ***setup_tone_curves(float curveatt_dB[P_BANDS],float binHz,int n,
87 float center_boost, float center_decay_rate){
90 float workc[P_BANDS][P_LEVELS][EHMER_MAX];
91 float athc[P_LEVELS][EHMER_MAX];
92 float *brute_buffer=alloca(n*sizeof(*brute_buffer));
94 float ***ret=_ogg_malloc(sizeof(*ret)*P_BANDS);
96 memset(workc,0,sizeof(workc));
98 for(i=0;i<P_BANDS;i++){
99 /* we add back in the ATH to avoid low level curves falling off to
100 -infinity and unnecessarily cutting off high level curves in the
101 curve limiting (last step). */
103 /* A half-band's settings must be valid over the whole band, and
104 it's better to mask too little than too much */
106 for(j=0;j<EHMER_MAX;j++){
109 if(j+k+ath_offset<MAX_ATH){
110 if(min>ATH[j+k+ath_offset])min=ATH[j+k+ath_offset];
112 if(min>ATH[MAX_ATH-1])min=ATH[MAX_ATH-1];
117 /* copy curves into working space, replicate the 50dB curve to 30
118 and 40, replicate the 100dB curve to 110 */
120 memcpy(workc[i][j+2],tonemasks[i][j],EHMER_MAX*sizeof(*tonemasks[i][j]));
121 memcpy(workc[i][0],tonemasks[i][0],EHMER_MAX*sizeof(*tonemasks[i][0]));
122 memcpy(workc[i][1],tonemasks[i][0],EHMER_MAX*sizeof(*tonemasks[i][0]));
124 /* apply centered curve boost/decay */
125 for(j=0;j<P_LEVELS;j++){
126 for(k=0;k<EHMER_MAX;k++){
127 float adj=center_boost+abs(EHMER_OFFSET-k)*center_decay_rate;
128 if(adj<0. && center_boost>0)adj=0.;
129 if(adj>0. && center_boost<0)adj=0.;
134 /* normalize curves so the driving amplitude is 0dB */
135 /* make temp curves with the ATH overlayed */
136 for(j=0;j<P_LEVELS;j++){
137 attenuate_curve(workc[i][j],curveatt_dB[i]+100.-(j<2?2:j)*10.-P_LEVEL_0);
138 memcpy(athc[j],ath,EHMER_MAX*sizeof(**athc));
139 attenuate_curve(athc[j],+100.-j*10.f-P_LEVEL_0);
140 max_curve(athc[j],workc[i][j]);
143 /* Now limit the louder curves.
145 the idea is this: We don't know what the playback attenuation
146 will be; 0dB SL moves every time the user twiddles the volume
147 knob. So that means we have to use a single 'most pessimal' curve
148 for all masking amplitudes, right? Wrong. The *loudest* sound
149 can be in (we assume) a range of ...+100dB] SL. However, sounds
150 20dB down will be in a range ...+80], 40dB down is from ...+60],
153 for(j=1;j<P_LEVELS;j++){
154 min_curve(athc[j],athc[j-1]);
155 min_curve(workc[i][j],athc[j]);
159 for(i=0;i<P_BANDS;i++){
160 int hi_curve,lo_curve,bin;
161 ret[i]=_ogg_malloc(sizeof(**ret)*P_LEVELS);
163 /* low frequency curves are measured with greater resolution than
164 the MDCT/FFT will actually give us; we want the curve applied
165 to the tone data to be pessimistic and thus apply the minimum
166 masking possible for a given bin. That means that a single bin
167 could span more than one octave and that the curve will be a
168 composite of multiple octaves. It also may mean that a single
169 bin may span > an eighth of an octave and that the eighth
170 octave values may also be composited. */
172 /* which octave curves will we be compositing? */
173 bin=floor(fromOC(i*.5)/binHz);
174 lo_curve= ceil(toOC(bin*binHz+1)*2);
175 hi_curve= floor(toOC((bin+1)*binHz)*2);
176 if(lo_curve>i)lo_curve=i;
177 if(lo_curve<0)lo_curve=0;
178 if(hi_curve>=P_BANDS)hi_curve=P_BANDS-1;
180 for(m=0;m<P_LEVELS;m++){
181 ret[i][m]=_ogg_malloc(sizeof(***ret)*(EHMER_MAX+2));
183 for(j=0;j<n;j++)brute_buffer[j]=999.;
185 /* render the curve into bins, then pull values back into curve.
186 The point is that any inherent subsampling aliasing results in
188 for(k=lo_curve;k<=hi_curve;k++){
191 for(j=0;j<EHMER_MAX;j++){
192 int lo_bin= fromOC(j*.125+k*.5-2.0625)/binHz;
193 int hi_bin= fromOC(j*.125+k*.5-1.9375)/binHz+1;
195 if(lo_bin<0)lo_bin=0;
196 if(lo_bin>n)lo_bin=n;
197 if(lo_bin<l)l=lo_bin;
198 if(hi_bin<0)hi_bin=0;
199 if(hi_bin>n)hi_bin=n;
201 for(;l<hi_bin && l<n;l++)
202 if(brute_buffer[l]>workc[k][m][j])
203 brute_buffer[l]=workc[k][m][j];
207 if(brute_buffer[l]>workc[k][m][EHMER_MAX-1])
208 brute_buffer[l]=workc[k][m][EHMER_MAX-1];
212 /* be equally paranoid about being valid up to next half ocatve */
216 for(j=0;j<EHMER_MAX;j++){
217 int lo_bin= fromOC(j*.125+i*.5-2.0625)/binHz;
218 int hi_bin= fromOC(j*.125+i*.5-1.9375)/binHz+1;
220 if(lo_bin<0)lo_bin=0;
221 if(lo_bin>n)lo_bin=n;
222 if(lo_bin<l)l=lo_bin;
223 if(hi_bin<0)hi_bin=0;
224 if(hi_bin>n)hi_bin=n;
226 for(;l<hi_bin && l<n;l++)
227 if(brute_buffer[l]>workc[k][m][j])
228 brute_buffer[l]=workc[k][m][j];
232 if(brute_buffer[l]>workc[k][m][EHMER_MAX-1])
233 brute_buffer[l]=workc[k][m][EHMER_MAX-1];
238 for(j=0;j<EHMER_MAX;j++){
239 int bin=fromOC(j*.125+i*.5-2.)/binHz;
241 ret[i][m][j+2]=-999.;
244 ret[i][m][j+2]=-999.;
246 ret[i][m][j+2]=brute_buffer[bin];
252 for(j=0;j<EHMER_OFFSET;j++)
253 if(ret[i][m][j+2]>-200.f)break;
256 for(j=EHMER_MAX-1;j>EHMER_OFFSET+1;j--)
257 if(ret[i][m][j+2]>-200.f)
267 void _vp_psy_init(vorbis_look_psy *p,vorbis_info_psy *vi,
268 vorbis_info_psy_global *gi,int n,long rate){
269 long i,j,lo=-99,hi=1;
271 memset(p,0,sizeof(*p));
273 p->eighth_octave_lines=gi->eighth_octave_lines;
274 p->shiftoc=rint(log(gi->eighth_octave_lines*8.f)/log(2.f))-1;
276 p->firstoc=toOC(.25f*rate*.5/n)*(1<<(p->shiftoc+1))-gi->eighth_octave_lines;
277 maxoc=toOC((n+.25f)*rate*.5/n)*(1<<(p->shiftoc+1))+.5f;
278 p->total_octave_lines=maxoc-p->firstoc+1;
279 p->ath=_ogg_malloc(n*sizeof(*p->ath));
281 p->octave=_ogg_malloc(n*sizeof(*p->octave));
282 p->bark=_ogg_malloc(n*sizeof(*p->bark));
287 /* AoTuV HF weighting */
289 if(rate < 26000) p->m_val = 0;
290 else if(rate < 38000) p->m_val = .94; /* 32kHz */
291 else if(rate > 46000) p->m_val = 1.275; /* 48kHz */
293 /* set up the lookups for a given blocksize and sample rate */
295 for(i=0,j=0;i<MAX_ATH-1;i++){
296 int endpos=rint(fromOC((i+1)*.125-2.)*2*n/rate);
299 float delta=(ATH[i+1]-base)/(endpos-j);
300 for(;j<endpos && j<n;j++){
308 p->ath[j]=p->ath[j-1];
312 float bark=toBARK(rate/(2*n)*i);
314 for(;lo+vi->noisewindowlomin<i &&
315 toBARK(rate/(2*n)*lo)<(bark-vi->noisewindowlo);lo++);
317 for(;hi<=n && (hi<i+vi->noisewindowhimin ||
318 toBARK(rate/(2*n)*hi)<(bark+vi->noisewindowhi));hi++);
320 p->bark[i]=((lo-1)<<16)+(hi-1);
325 p->octave[i]=toOC((i+.25f)*.5*rate/n)*(1<<(p->shiftoc+1))+.5f;
327 p->tonecurves=setup_tone_curves(vi->toneatt,rate*.5/n,n,
328 vi->tone_centerboost,vi->tone_decay);
330 /* set up rolling noise median */
331 p->noiseoffset=_ogg_malloc(P_NOISECURVES*sizeof(*p->noiseoffset));
332 for(i=0;i<P_NOISECURVES;i++)
333 p->noiseoffset[i]=_ogg_malloc(n*sizeof(**p->noiseoffset));
336 float halfoc=toOC((i+.5)*rate/(2.*n))*2.;
340 if(halfoc<0)halfoc=0;
341 if(halfoc>=P_BANDS-1)halfoc=P_BANDS-1;
342 inthalfoc=(int)halfoc;
343 del=halfoc-inthalfoc;
345 for(j=0;j<P_NOISECURVES;j++)
346 p->noiseoffset[j][i]=
347 p->vi->noiseoff[j][inthalfoc]*(1.-del) +
348 p->vi->noiseoff[j][inthalfoc+1]*del;
354 _analysis_output_always("noiseoff0",ls,p->noiseoffset[0],n,1,0,0);
355 _analysis_output_always("noiseoff1",ls,p->noiseoffset[1],n,1,0,0);
356 _analysis_output_always("noiseoff2",ls++,p->noiseoffset[2],n,1,0,0);
361 void _vp_psy_clear(vorbis_look_psy *p){
364 if(p->ath)_ogg_free(p->ath);
365 if(p->octave)_ogg_free(p->octave);
366 if(p->bark)_ogg_free(p->bark);
368 for(i=0;i<P_BANDS;i++){
369 for(j=0;j<P_LEVELS;j++){
370 _ogg_free(p->tonecurves[i][j]);
372 _ogg_free(p->tonecurves[i]);
374 _ogg_free(p->tonecurves);
377 for(i=0;i<P_NOISECURVES;i++){
378 _ogg_free(p->noiseoffset[i]);
380 _ogg_free(p->noiseoffset);
382 memset(p,0,sizeof(*p));
386 /* octave/(8*eighth_octave_lines) x scale and dB y scale */
387 static void seed_curve(float *seed,
388 const float **curves,
391 int linesper,float dBoffset){
394 const float *posts,*curve;
396 int choice=(int)((amp+dBoffset-P_LEVEL_0)*.1f);
397 choice=max(choice,0);
398 choice=min(choice,P_LEVELS-1);
399 posts=curves[choice];
402 seedptr=oc+(posts[0]-EHMER_OFFSET)*linesper-(linesper>>1);
404 for(i=posts[0];i<post1;i++){
406 float lin=amp+curve[i];
407 if(seed[seedptr]<lin)seed[seedptr]=lin;
414 static void seed_loop(vorbis_look_psy *p,
415 const float ***curves,
420 vorbis_info_psy *vi=p->vi;
422 float dBoffset=vi->max_curve_dB-specmax;
424 /* prime the working vector with peak values */
428 long oc=p->octave[i];
429 while(i+1<n && p->octave[i+1]==oc){
431 if(f[i]>max)max=f[i];
437 if(oc>=P_BANDS)oc=P_BANDS-1;
443 p->octave[i]-p->firstoc,
444 p->total_octave_lines,
445 p->eighth_octave_lines,
451 static void seed_chase(float *seeds, int linesper, long n){
452 long *posstack=alloca(n*sizeof(*posstack));
453 float *ampstack=alloca(n*sizeof(*ampstack));
461 ampstack[stack++]=seeds[i];
464 if(seeds[i]<ampstack[stack-1]){
466 ampstack[stack++]=seeds[i];
469 if(i<posstack[stack-1]+linesper){
470 if(stack>1 && ampstack[stack-1]<=ampstack[stack-2] &&
471 i<posstack[stack-2]+linesper){
472 /* we completely overlap, making stack-1 irrelevant. pop it */
478 ampstack[stack++]=seeds[i];
486 /* the stack now contains only the positions that are relevant. Scan
487 'em straight through */
489 for(i=0;i<stack;i++){
491 if(i<stack-1 && ampstack[i+1]>ampstack[i]){
492 endpos=posstack[i+1];
494 endpos=posstack[i]+linesper+1; /* +1 is important, else bin 0 is
495 discarded in short frames */
497 if(endpos>n)endpos=n;
498 for(;pos<endpos;pos++)
499 seeds[pos]=ampstack[i];
502 /* there. Linear time. I now remember this was on a problem set I
503 had in Grad Skool... I didn't solve it at the time ;-) */
507 /* bleaugh, this is more complicated than it needs to be */
509 static void max_seeds(vorbis_look_psy *p,
512 long n=p->total_octave_lines;
513 int linesper=p->eighth_octave_lines;
517 seed_chase(seed,linesper,n); /* for masking */
519 pos=p->octave[0]-p->firstoc-(linesper>>1);
521 while(linpos+1<p->n){
522 float minV=seed[pos];
523 long end=((p->octave[linpos]+p->octave[linpos+1])>>1)-p->firstoc;
524 if(minV>p->vi->tone_abs_limit)minV=p->vi->tone_abs_limit;
527 if((seed[pos]>NEGINF && seed[pos]<minV) || minV==NEGINF)
532 for(;linpos<p->n && p->octave[linpos]<=end;linpos++)
533 if(flr[linpos]<minV)flr[linpos]=minV;
537 float minV=seed[p->total_octave_lines-1];
538 for(;linpos<p->n;linpos++)
539 if(flr[linpos]<minV)flr[linpos]=minV;
544 static void bark_noise_hybridmp(int n,const long *b,
550 float *N=alloca(n*sizeof(*N));
551 float *X=alloca(n*sizeof(*N));
552 float *XX=alloca(n*sizeof(*N));
553 float *Y=alloca(n*sizeof(*N));
554 float *XY=alloca(n*sizeof(*N));
556 float tN, tX, tXX, tY, tXY;
566 tN = tX = tXX = tY = tXY = 0.f;
569 if (y < 1.f) y = 1.f;
583 for (i = 1, x = 1.f; i < n; i++, x += 1.f) {
586 if (y < 1.f) y = 1.f;
603 for (i = 0, x = 0.f;; i++, x += 1.f) {
611 tXX = XX[hi] + XX[-lo];
613 tXY = XY[hi] - XY[-lo];
615 A = tY * tXX - tX * tXY;
616 B = tN * tXY - tX * tY;
617 D = tN * tXX - tX * tX;
622 noise[i] = R - offset;
625 for ( ;; i++, x += 1.f) {
633 tXX = XX[hi] - XX[lo];
635 tXY = XY[hi] - XY[lo];
637 A = tY * tXX - tX * tXY;
638 B = tN * tXY - tX * tY;
639 D = tN * tXX - tX * tX;
641 if (R < 0.f) R = 0.f;
643 noise[i] = R - offset;
645 for ( ; i < n; i++, x += 1.f) {
648 if (R < 0.f) R = 0.f;
650 noise[i] = R - offset;
653 if (fixed <= 0) return;
655 for (i = 0, x = 0.f;; i++, x += 1.f) {
662 tXX = XX[hi] + XX[-lo];
664 tXY = XY[hi] - XY[-lo];
667 A = tY * tXX - tX * tXY;
668 B = tN * tXY - tX * tY;
669 D = tN * tXX - tX * tX;
672 if (R - offset < noise[i]) noise[i] = R - offset;
674 for ( ;; i++, x += 1.f) {
682 tXX = XX[hi] - XX[lo];
684 tXY = XY[hi] - XY[lo];
686 A = tY * tXX - tX * tXY;
687 B = tN * tXY - tX * tY;
688 D = tN * tXX - tX * tX;
691 if (R - offset < noise[i]) noise[i] = R - offset;
693 for ( ; i < n; i++, x += 1.f) {
695 if (R - offset < noise[i]) noise[i] = R - offset;
699 static const float FLOOR1_fromdB_INV_LOOKUP[256]={
700 0.F, 8.81683e+06F, 8.27882e+06F, 7.77365e+06F,
701 7.29930e+06F, 6.85389e+06F, 6.43567e+06F, 6.04296e+06F,
702 5.67422e+06F, 5.32798e+06F, 5.00286e+06F, 4.69759e+06F,
703 4.41094e+06F, 4.14178e+06F, 3.88905e+06F, 3.65174e+06F,
704 3.42891e+06F, 3.21968e+06F, 3.02321e+06F, 2.83873e+06F,
705 2.66551e+06F, 2.50286e+06F, 2.35014e+06F, 2.20673e+06F,
706 2.07208e+06F, 1.94564e+06F, 1.82692e+06F, 1.71544e+06F,
707 1.61076e+06F, 1.51247e+06F, 1.42018e+06F, 1.33352e+06F,
708 1.25215e+06F, 1.17574e+06F, 1.10400e+06F, 1.03663e+06F,
709 973377.F, 913981.F, 858210.F, 805842.F,
710 756669.F, 710497.F, 667142.F, 626433.F,
711 588208.F, 552316.F, 518613.F, 486967.F,
712 457252.F, 429351.F, 403152.F, 378551.F,
713 355452.F, 333762.F, 313396.F, 294273.F,
714 276316.F, 259455.F, 243623.F, 228757.F,
715 214798.F, 201691.F, 189384.F, 177828.F,
716 166977.F, 156788.F, 147221.F, 138237.F,
717 129802.F, 121881.F, 114444.F, 107461.F,
718 100903.F, 94746.3F, 88964.9F, 83536.2F,
719 78438.8F, 73652.5F, 69158.2F, 64938.1F,
720 60975.6F, 57254.9F, 53761.2F, 50480.6F,
721 47400.3F, 44507.9F, 41792.0F, 39241.9F,
722 36847.3F, 34598.9F, 32487.7F, 30505.3F,
723 28643.8F, 26896.0F, 25254.8F, 23713.7F,
724 22266.7F, 20908.0F, 19632.2F, 18434.2F,
725 17309.4F, 16253.1F, 15261.4F, 14330.1F,
726 13455.7F, 12634.6F, 11863.7F, 11139.7F,
727 10460.0F, 9821.72F, 9222.39F, 8659.64F,
728 8131.23F, 7635.06F, 7169.17F, 6731.70F,
729 6320.93F, 5935.23F, 5573.06F, 5232.99F,
730 4913.67F, 4613.84F, 4332.30F, 4067.94F,
731 3819.72F, 3586.64F, 3367.78F, 3162.28F,
732 2969.31F, 2788.13F, 2617.99F, 2458.24F,
733 2308.24F, 2167.39F, 2035.14F, 1910.95F,
734 1794.35F, 1684.85F, 1582.04F, 1485.51F,
735 1394.86F, 1309.75F, 1229.83F, 1154.78F,
736 1084.32F, 1018.15F, 956.024F, 897.687F,
737 842.910F, 791.475F, 743.179F, 697.830F,
738 655.249F, 615.265F, 577.722F, 542.469F,
739 509.367F, 478.286F, 449.101F, 421.696F,
740 395.964F, 371.803F, 349.115F, 327.812F,
741 307.809F, 289.026F, 271.390F, 254.830F,
742 239.280F, 224.679F, 210.969F, 198.096F,
743 186.008F, 174.658F, 164.000F, 153.993F,
744 144.596F, 135.773F, 127.488F, 119.708F,
745 112.404F, 105.545F, 99.1046F, 93.0572F,
746 87.3788F, 82.0469F, 77.0404F, 72.3394F,
747 67.9252F, 63.7804F, 59.8885F, 56.2341F,
748 52.8027F, 49.5807F, 46.5553F, 43.7144F,
749 41.0470F, 38.5423F, 36.1904F, 33.9821F,
750 31.9085F, 29.9614F, 28.1332F, 26.4165F,
751 24.8045F, 23.2910F, 21.8697F, 20.5352F,
752 19.2822F, 18.1056F, 17.0008F, 15.9634F,
753 14.9893F, 14.0746F, 13.2158F, 12.4094F,
754 11.6522F, 10.9411F, 10.2735F, 9.64662F,
755 9.05798F, 8.50526F, 7.98626F, 7.49894F,
756 7.04135F, 6.61169F, 6.20824F, 5.82941F,
757 5.47370F, 5.13970F, 4.82607F, 4.53158F,
758 4.25507F, 3.99542F, 3.75162F, 3.52269F,
759 3.30774F, 3.10590F, 2.91638F, 2.73842F,
760 2.57132F, 2.41442F, 2.26709F, 2.12875F,
761 1.99885F, 1.87688F, 1.76236F, 1.65482F,
762 1.55384F, 1.45902F, 1.36999F, 1.28640F,
763 1.20790F, 1.13419F, 1.06499F, 1.F
766 void _vp_remove_floor(vorbis_look_psy *p,
770 int sliding_lowpass){
774 if(sliding_lowpass>n)sliding_lowpass=n;
776 for(i=0;i<sliding_lowpass;i++){
778 mdct[i]*FLOOR1_fromdB_INV_LOOKUP[codedflr[i]];
785 void _vp_noisemask(vorbis_look_psy *p,
790 float *work=alloca(n*sizeof(*work));
792 bark_noise_hybridmp(n,p->bark,logmdct,logmask,
795 for(i=0;i<n;i++)work[i]=logmdct[i]-logmask[i];
797 bark_noise_hybridmp(n,p->bark,work,logmask,0.,
798 p->vi->noisewindowfixed);
800 for(i=0;i<n;i++)work[i]=logmdct[i]-work[i];
808 work2[i]=logmask[i]+work[i];
812 _analysis_output("median2R",seq/2,work,n,1,0,0);
814 _analysis_output("median2L",seq/2,work,n,1,0,0);
817 _analysis_output("envelope2R",seq/2,work2,n,1,0,0);
819 _analysis_output("envelope2L",seq/2,work2,n,1,0,0);
825 int dB=logmask[i]+.5;
826 if(dB>=NOISE_COMPAND_LEVELS)dB=NOISE_COMPAND_LEVELS-1;
828 logmask[i]= work[i]+p->vi->noisecompand[dB];
833 void _vp_tonemask(vorbis_look_psy *p,
836 float global_specmax,
837 float local_specmax){
841 float *seed=alloca(sizeof(*seed)*p->total_octave_lines);
842 float att=local_specmax+p->vi->ath_adjatt;
843 for(i=0;i<p->total_octave_lines;i++)seed[i]=NEGINF;
845 /* set the ATH (floating below localmax, not global max by a
847 if(att<p->vi->ath_maxatt)att=p->vi->ath_maxatt;
850 logmask[i]=p->ath[i]+att;
853 seed_loop(p,(const float ***)p->tonecurves,logfft,logmask,seed,global_specmax);
854 max_seeds(p,seed,logmask);
858 void _vp_offset_and_mix(vorbis_look_psy *p,
866 float de, coeffi, cx;/* AoTuV */
867 float toneatt=p->vi->tone_masteratt[offset_select];
872 float val= noise[i]+p->noiseoffset[offset_select][i];
873 if(val>p->vi->noisemaxsupp)val=p->vi->noisemaxsupp;
874 logmask[i]=max(val,tone[i]+toneatt);
879 The following codes improve a noise problem.
880 A fundamental idea uses the value of masking and carries out
881 the relative compensation of the MDCT.
882 However, this code is not perfect and all noise problems cannot be solved.
883 by Aoyumi @ 2004/04/18
886 if(offset_select == 1) {
887 coeffi = -17.2; /* coeffi is a -17.2dB threshold */
888 val = val - logmdct[i]; /* val == mdct line value relative to floor in dB */
891 /* mdct value is > -17.2 dB below floor */
893 de = 1.0-((val-coeffi)*0.005*cx);
894 /* pro-rated attenuation:
895 -0.00 dB boost if mdct value is -17.2dB (relative to floor)
896 -0.77 dB boost if mdct value is 0dB (relative to floor)
897 -1.64 dB boost if mdct value is +17.2dB (relative to floor)
900 if(de < 0) de = 0.0001;
902 /* mdct value is <= -17.2 dB below floor */
904 de = 1.0-((val-coeffi)*0.0003*cx);
905 /* pro-rated attenuation:
906 +0.00 dB atten if mdct value is -17.2dB (relative to floor)
907 +0.45 dB atten if mdct value is -34.4dB (relative to floor)
916 float _vp_ampmax_decay(float amp,vorbis_dsp_state *vd){
917 vorbis_info *vi=vd->vi;
918 codec_setup_info *ci=vi->codec_setup;
919 vorbis_info_psy_global *gi=&ci->psy_g_param;
921 int n=ci->blocksizes[vd->W]/2;
922 float secs=(float)n/vi->rate;
924 amp+=secs*gi->ampmax_att_per_sec;
925 if(amp<-9999)amp=-9999;
929 static void couple_lossless(float A, float B,
930 float *qA, float *qB){
931 int test1=fabs(*qA)>fabs(*qB);
932 test1-= fabs(*qA)<fabs(*qB);
934 //if(!test1)test1=((fabs(A)>fabs(B))<<1)-1;
936 *qB=(*qA>0.f?*qA-*qB:*qB-*qA);
939 *qB=(*qB>0.f?*qA-*qB:*qB-*qA);
943 if(*qB>fabs(*qA)*1.9999f){
949 static const float hypot_lookup[32]={
950 -0.009935, -0.011245, -0.012726, -0.014397,
951 -0.016282, -0.018407, -0.020800, -0.023494,
952 -0.026522, -0.029923, -0.033737, -0.038010,
953 -0.042787, -0.048121, -0.054064, -0.060671,
954 -0.068000, -0.076109, -0.085054, -0.094892,
955 -0.105675, -0.117451, -0.130260, -0.144134,
956 -0.159093, -0.175146, -0.192286, -0.210490,
957 -0.229718, -0.249913, -0.271001, -0.292893};
959 static void precomputed_couple_point(float premag,
960 int floorA,int floorB,
961 float *mag, float *ang){
963 int test=(floorA>floorB)-1;
964 int offset=31-abs(floorA-floorB);
965 float floormag=hypot_lookup[((offset<0)-1)&offset]+1.f;
967 floormag*=FLOOR1_fromdB_INV_LOOKUP[(floorB&test)|(floorA&(~test))];
969 *mag=premag*floormag;
973 /* just like below, this is currently set up to only do
974 single-step-depth coupling. Otherwise, we'd have to do more
975 copying (which will be inevitable later) */
977 /* doing the real circular magnitude calculation is audibly superior
979 static float dipole_hypot(float a, float b){
981 if(b>0.)return sqrt(a*a+b*b);
982 if(a>-b)return sqrt(a*a-b*b);
983 return -sqrt(b*b-a*a);
985 if(b<0.)return -sqrt(a*a+b*b);
986 if(-a>b)return -sqrt(a*a-b*b);
987 return sqrt(b*b-a*a);
989 static float round_hypot(float a, float b){
991 if(b>0.)return sqrt(a*a+b*b);
992 if(a>-b)return sqrt(a*a+b*b);
993 return -sqrt(b*b+a*a);
995 if(b<0.)return -sqrt(a*a+b*b);
996 if(-a>b)return -sqrt(a*a+b*b);
997 return sqrt(b*b+a*a);
1000 /* revert to round hypot for now */
1001 float **_vp_quantize_couple_memo(vorbis_block *vb,
1002 vorbis_info_psy_global *g,
1004 vorbis_info_mapping0 *vi,
1008 float **ret=_vorbis_block_alloc(vb,vi->coupling_steps*sizeof(*ret));
1009 int limit=g->coupling_pointlimit[p->vi->blockflag][PACKETBLOBS/2];
1011 for(i=0;i<vi->coupling_steps;i++){
1012 float *mdctM=mdct[vi->coupling_mag[i]];
1013 float *mdctA=mdct[vi->coupling_ang[i]];
1014 ret[i]=_vorbis_block_alloc(vb,n*sizeof(**ret));
1015 for(j=0;j<limit;j++)
1016 ret[i][j]=dipole_hypot(mdctM[j],mdctA[j]);
1018 ret[i][j]=round_hypot(mdctM[j],mdctA[j]);
1024 /* this is for per-channel noise normalization */
1025 static int apsort(const void *a, const void *b){
1026 float f1=fabs(**(float**)a);
1027 float f2=fabs(**(float**)b);
1028 return (f1<f2)-(f1>f2);
1031 int **_vp_quantize_couple_sort(vorbis_block *vb,
1033 vorbis_info_mapping0 *vi,
1037 if(p->vi->normal_point_p){
1039 int **ret=_vorbis_block_alloc(vb,vi->coupling_steps*sizeof(*ret));
1040 int partition=p->vi->normal_partition;
1041 float **work=alloca(sizeof(*work)*partition);
1043 for(i=0;i<vi->coupling_steps;i++){
1044 ret[i]=_vorbis_block_alloc(vb,n*sizeof(**ret));
1046 for(j=0;j<n;j+=partition){
1047 for(k=0;k<partition;k++)work[k]=mags[i]+k+j;
1048 qsort(work,partition,sizeof(*work),apsort);
1049 for(k=0;k<partition;k++)ret[i][k+j]=work[k]-mags[i];
1057 void _vp_noise_normalize_sort(vorbis_look_psy *p,
1058 float *magnitudes,int *sortedindex){
1060 vorbis_info_psy *vi=p->vi;
1061 int partition=vi->normal_partition;
1062 float **work=alloca(sizeof(*work)*partition);
1063 int start=vi->normal_start;
1065 for(j=start;j<n;j+=partition){
1066 if(j+partition>n)partition=n-j;
1067 for(i=0;i<partition;i++)work[i]=magnitudes+i+j;
1068 qsort(work,partition,sizeof(*work),apsort);
1069 for(i=0;i<partition;i++){
1070 sortedindex[i+j-start]=work[i]-magnitudes;
1075 void _vp_noise_normalize(vorbis_look_psy *p,
1076 float *in,float *out,int *sortedindex){
1077 int flag=0,i,j=0,n=p->n;
1078 vorbis_info_psy *vi=p->vi;
1079 int partition=vi->normal_partition;
1080 int start=vi->normal_start;
1084 if(vi->normal_channel_p){
1088 for(;j+partition<=n;j+=partition){
1092 for(i=j;i<j+partition;i++)
1095 for(i=0;i<partition;i++){
1096 k=sortedindex[i+j-start];
1098 if(in[k]*in[k]>=.25f){
1103 if(acc<vi->normal_thresh)break;
1104 out[k]=unitnorm(in[k]);
1109 for(;i<partition;i++){
1110 k=sortedindex[i+j-start];
1121 void _vp_couple(int blobno,
1122 vorbis_info_psy_global *g,
1124 vorbis_info_mapping0 *vi,
1130 int sliding_lowpass){
1134 /* perform any requested channel coupling */
1135 /* point stereo can only be used in a first stage (in this encoder)
1136 because of the dependency on floor lookups */
1137 for(i=0;i<vi->coupling_steps;i++){
1139 /* once we're doing multistage coupling in which a channel goes
1140 through more than one coupling step, the floor vector
1141 magnitudes will also have to be recalculated an propogated
1142 along with PCM. Right now, we're not (that will wait until 5.1
1143 most likely), so the code isn't here yet. The memory management
1144 here is all assuming single depth couplings anyway. */
1146 /* make sure coupling a zero and a nonzero channel results in two
1147 nonzero channels. */
1148 if(nonzero[vi->coupling_mag[i]] ||
1149 nonzero[vi->coupling_ang[i]]){
1151 float *rM=res[vi->coupling_mag[i]];
1152 float *rA=res[vi->coupling_ang[i]];
1155 int *floorM=ifloor[vi->coupling_mag[i]];
1156 int *floorA=ifloor[vi->coupling_ang[i]];
1157 float prepoint=stereo_threshholds[g->coupling_prepointamp[blobno]];
1158 float postpoint=stereo_threshholds[g->coupling_postpointamp[blobno]];
1159 int partition=(p->vi->normal_point_p?p->vi->normal_partition:p->n);
1160 int limit=g->coupling_pointlimit[p->vi->blockflag][blobno];
1162 nonzero[vi->coupling_mag[i]]=1;
1163 nonzero[vi->coupling_ang[i]]=1;
1165 /* The threshold of a stereo is changed with the size of n */
1167 postpoint=stereo_threshholds_limited[g->coupling_postpointamp[blobno]];
1169 for(j=0;j<p->n;j+=partition){
1172 for(k=0;k<partition;k++){
1175 if(l<sliding_lowpass){
1176 if((l>=limit && fabs(rM[l])<postpoint && fabs(rA[l])<postpoint) ||
1177 (fabs(rM[l])<prepoint && fabs(rA[l])<prepoint)){
1179 precomputed_couple_point(mag_memo[i][l],
1180 floorM[l],floorA[l],
1183 if(rint(qM[l])==0.f)acc+=qM[l]*qM[l];
1185 couple_lossless(rM[l],rA[l],qM+l,qA+l);
1193 if(p->vi->normal_point_p){
1194 for(k=0;k<partition && acc>=p->vi->normal_thresh;k++){
1195 int l=mag_sort[i][j+k];
1196 if(l<sliding_lowpass && l>=limit && rint(qM[l])==0.f){
1197 qM[l]=unitnorm(qM[l]);
1209 The boost problem by the combination of noise normalization and point stereo is eased.
1210 However, this is a temporary patch.
1211 by Aoyumi @ 2004/04/18
1214 void hf_reduction(vorbis_info_psy_global *g,
1216 vorbis_info_mapping0 *vi,
1219 int i,j,n=p->n, de=0.3*p->m_val;
1220 int limit=g->coupling_pointlimit[p->vi->blockflag][PACKETBLOBS/2];
1222 for(i=0; i<vi->coupling_steps; i++){
1223 /* for(j=start; j<limit; j++){} // ???*/
1224 for(j=limit; j<n; j++)
1225 mdct[i][j] *= (1.0 - de*((float)(j-limit) / (float)(n-limit)));