lib/envelope.c

   1 /********************************************************************
   2  *                                                                  *
   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.       *
   7  *                                                                  *
   8  * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2001             *
   9  * by the XIPHOPHORUS Company http://www.xiph.org/                  *
  10  *                                                                  *
  11  ********************************************************************
  12
  13  function: PCM data envelope analysis
  14  last mod: $Id: envelope.c,v 1.46 2002/03/29 08:18:23 xiphmont Exp $
  15
  16  ********************************************************************/
  17
  18 #include <stdlib.h>
  19 #include <string.h>
  20 #include <stdio.h>
  21 #include <math.h>
  22 #include <ogg/ogg.h>
  23 #include "vorbis/codec.h"
  24 #include "codec_internal.h"
  25
  26 #include "os.h"
  27 #include "scales.h"
  28 #include "envelope.h"
  29 #include "mdct.h"
  30 #include "misc.h"
  31
  32 void _ve_envelope_init(envelope_lookup *e,vorbis_info *vi){
  33   codec_setup_info *ci=vi->codec_setup;
  34   vorbis_info_psy_global *gi=&ci->psy_g_param;
  35   int ch=vi->channels;
  36   int i,j;
  37   int n=e->winlength=128;
  38   e->searchstep=64; /* not random */
  39
  40   e->minenergy=gi->preecho_minenergy;
  41   e->ch=ch;
  42   e->storage=128;
  43   e->cursor=ci->blocksizes[1]/2;
  44   e->mdct_win=_ogg_calloc(n,sizeof(*e->mdct_win));
  45   mdct_init(&e->mdct,n);
  46
  47   for(i=0;i<n;i++){
  48     e->mdct_win[i]=sin(i/(n-1.)*M_PI);
  49     e->mdct_win[i]*=e->mdct_win[i];
  50   }
  51
  52   /* magic follows */
  53   e->band[0].begin=4;  e->band[0].end=8;
  54   e->band[1].begin=8;  e->band[1].end=10;
  55   e->band[2].begin=14; e->band[2].end=12;
  56   e->band[3].begin=20; e->band[3].end=16;
  57   e->band[4].begin=28; e->band[4].end=20;
  58   e->band[5].begin=40; e->band[5].end=20;
  59
  60   for(j=0;j<VE_BANDS;j++){
  61     n=e->band[j].end;
  62     e->band[j].window=_ogg_malloc(n*sizeof(*e->band[0].window));
  63     for(i=0;i<n;i++){
  64       e->band[j].window[i]=sin((i+.5)/n*M_PI);
  65       e->band[j].total+=e->band[j].window[i];
  66     }
  67     e->band[j].total=1./e->band[j].total;
  68   }
  69
  70   e->filter=_ogg_calloc(VE_BANDS*ch,sizeof(*e->filter));
  71   e->mark=_ogg_calloc(e->storage,sizeof(*e->mark));
  72
  73 }
  74
  75 void _ve_envelope_clear(envelope_lookup *e){
  76   int i;
  77   mdct_clear(&e->mdct);
  78   for(i=0;i<VE_BANDS;i++)
  79     _ogg_free(e->band[i].window);
  80   _ogg_free(e->mdct_win);
  81   _ogg_free(e->filter);
  82   _ogg_free(e->mark);
  83   memset(e,0,sizeof(*e));
  84 }
  85
  86 extern void _analysis_output_always(char *base,int i,float *v,int n,int bark,int dB,ogg_int64_t off);
  87
  88 /* fairly straight threshhold-by-band based until we find something
  89    that works better and isn't patented. */
  90
  91 static int _ve_amp(envelope_lookup *ve,
  92                    vorbis_info_psy_global *gi,
  93                    float *data,
  94                    envelope_band *bands,
  95                    envelope_filter_state *filters,
  96                    long pos){
  97   long n=ve->winlength;
  98   int ret=0;
  99   long i,j;
 100   float decay;
 101
 102   /* we want to have a 'minimum bar' for energy, else we're just
 103      basing blocks on quantization noise that outweighs the signal
 104      itself (for low power signals) */
 105
 106   float minV=ve->minenergy;
 107   float *vec=alloca(n*sizeof(*vec));
 108
 109   /* stretch is used to gradually lengthen the number of windows
 110      considered prevoius-to-potential-trigger */
 111   int stretch=max(VE_MINSTRETCH,ve->stretch/2);
 112   float penalty=gi->stretch_penalty-(ve->stretch/2-VE_MINSTRETCH);
 113   if(penalty<0.f)penalty=0.f;
 114   if(penalty>gi->stretch_penalty)penalty=gi->stretch_penalty;
 115
 116   /*_analysis_output_always("lpcm",seq2,data,n,0,0,
 117     totalshift+pos*ve->searchstep);*/
 118
 119  /* window and transform */
 120   for(i=0;i<n;i++)
 121     vec[i]=data[i]*ve->mdct_win[i];
 122   mdct_forward(&ve->mdct,vec,vec);
 123
 124   /* _analysis_output_always("mdct",seq2,vec,n/2,0,1,0); */
 125
 126   /* near-DC spreading function; this has nothing to do with
 127      psychoacoustics, just sidelobe leakage and window size */
 128   {
 129     float temp=vec[0]*vec[0]+.7*vec[1]*vec[1]+.2*vec[2]*vec[2];
 130     int ptr=filters->nearptr;
 131
 132     /* the accumulation is regularly refreshed from scratch to avoid
 133        floating point creep */
 134     if(ptr==0){
 135       decay=filters->nearDC_acc=filters->nearDC_partialacc+temp;
 136       filters->nearDC_partialacc=temp;
 137     }else{
 138       decay=filters->nearDC_acc+=temp;
 139       filters->nearDC_partialacc+=temp;
 140     }
 141     filters->nearDC_acc-=filters->nearDC[ptr];
 142     filters->nearDC[ptr]=temp;
 143
 144     decay*=(1./(VE_NEARDC+1));
 145     filters->nearptr++;
 146     if(filters->nearptr>=VE_NEARDC)filters->nearptr=0;
 147     decay=todB(&decay)*.5-15.f;
 148   }
 149
 150   /* perform spreading and limiting, also smooth the spectrum.  yes,
 151      the MDCT results in all real coefficients, but it still *behaves*
 152      like real/imaginary pairs */
 153   for(i=0;i<n/2;i+=2){
 154     float val=vec[i]*vec[i]+vec[i+1]*vec[i+1];
 155     val=todB(&val)*.5f;
 156     if(val<decay)val=decay;
 157     if(val<minV)val=minV;
 158     vec[i>>1]=val;
 159     decay-=8.;
 160   }
 161
 162   /* _analysis_output_always("spread",seq2++,vec,n/4,0,0,0);*/
 163
 164   /* perform preecho/postecho triggering by band */
 165   for(j=0;j<VE_BANDS;j++){
 166     float acc=0.;
 167     float valmax,valmin;
 168
 169     /* accumulate amplitude */
 170     for(i=0;i<bands[j].end;i++)
 171       acc+=vec[i+bands[j].begin]*bands[j].window[i];
 172
 173     acc*=bands[j].total;
 174
 175     /* convert amplitude to delta */
 176     {
 177       int p,this=filters[j].ampptr;
 178       float postmax,postmin,premax=-99999.f,premin=99999.f;
 179
 180       p=this;
 181       p--;
 182       if(p<0)p+=VE_AMP;
 183       postmax=max(acc,filters[j].ampbuf[p]);
 184       postmin=min(acc,filters[j].ampbuf[p]);
 185
 186       for(i=0;i<stretch;i++){
 187         p--;
 188         if(p<0)p+=VE_AMP;
 189         premax=max(premax,filters[j].ampbuf[p]);
 190         premin=min(premin,filters[j].ampbuf[p]);
 191       }
 192
 193       valmin=postmin-premin;
 194       valmax=postmax-premax;
 195
 196       filters[j].ampbuf[this]=acc;
 197       filters[j].ampptr++;
 198       if(filters[j].ampptr>=VE_AMP)filters[j].ampptr=0;
 199     }
 200
 201     /* look at min/max, decide trigger */
 202     if(valmax>gi->preecho_thresh[j]+penalty)ret|=1;
 203     if(valmin<gi->postecho_thresh[j]-penalty)ret|=2;
 204   }
 205
 206   if(ret&1)ve->stretch=-1;
 207
 208   return(ret);
 209 }
 210
 211 long _ve_envelope_search(vorbis_dsp_state *v){
 212   vorbis_info *vi=v->vi;
 213   codec_setup_info *ci=vi->codec_setup;
 214   vorbis_info_psy_global *gi=&ci->psy_g_param;
 215   envelope_lookup *ve=((backend_lookup_state *)(v->backend_state))->ve;
 216   long i,j;
 217
 218   int first=ve->current/ve->searchstep;
 219   int last=v->pcm_current/ve->searchstep-VE_WIN;
 220   if(first<0)first=0;
 221
 222   /* make sure we have enough storage to match the PCM */
 223   if(last>ve->storage){
 224     ve->storage=last+VE_WIN;
 225     ve->mark=_ogg_realloc(ve->mark,ve->storage*sizeof(*ve->mark));
 226   }
 227
 228   for(j=first;j<last;j++){
 229     int ret=0;
 230
 231     ve->stretch++;
 232     if(ve->stretch>VE_MAXSTRETCH*2)
 233       ve->stretch=VE_MAXSTRETCH*2;
 234
 235     for(i=0;i<ve->ch;i++){
 236       float *pcm=v->pcm[i]+ve->searchstep*(j);
 237       ret|=_ve_amp(ve,gi,pcm,ve->band,ve->filter+i*VE_BANDS,j);
 238     }
 239
 240     ve->mark[j+VE_POST]=0;
 241     if(ret&1){
 242       ve->mark[j]=1;
 243       ve->mark[j+1]=1;
 244     }
 245
 246     if(ret&2){
 247       ve->mark[j]=1;
 248       if(j>0)ve->mark[j-1]=1;
 249     }
 250   }
 251
 252   ve->current=last*ve->searchstep;
 253
 254   {
 255     long centerW=v->centerW;
 256     long testW=
 257       centerW+
 258       ci->blocksizes[v->W]/4+
 259       ci->blocksizes[1]/2+
 260       ci->blocksizes[0]/4;
 261
 262     j=ve->cursor;
 263
 264     while(j<ve->current-(ve->searchstep)){/* account for postecho
 265                                              working back one window */
 266       if(j>=testW)return(1);
 267
 268       ve->cursor=j;
 269
 270       if(ve->mark[j/ve->searchstep]){
 271         if(j>centerW){
 272
 273 #if 0
 274           if(j>ve->curmark){
 275             float *marker=alloca(v->pcm_current*sizeof(*marker));
 276             int l,m;
 277             memset(marker,0,sizeof(*marker)*v->pcm_current);
 278             fprintf(stderr,"mark! seq=%d, cursor:%fs time:%fs\n",
 279                     seq,
 280                     (totalshift+ve->cursor)/44100.,
 281                     (totalshift+j)/44100.);
 282             _analysis_output_always("pcmL",seq,v->pcm[0],v->pcm_current,0,0,totalshift);
 283             _analysis_output_always("pcmR",seq,v->pcm[1],v->pcm_current,0,0,totalshift);
 284
 285             _analysis_output_always("markL",seq,v->pcm[0],j,0,0,totalshift);
 286             _analysis_output_always("markR",seq,v->pcm[1],j,0,0,totalshift);
 287
 288             for(m=0;m<VE_BANDS;m++){
 289               char buf[80];
 290               sprintf(buf,"delL%d",m);
 291               for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[m].markers[l]*.1;
 292               _analysis_output_always(buf,seq,marker,v->pcm_current,0,0,totalshift);
 293             }
 294
 295             for(m=0;m<VE_BANDS;m++){
 296               char buf[80];
 297               sprintf(buf,"delR%d",m);
 298               for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[m+VE_BANDS].markers[l]*.1;
 299               _analysis_output_always(buf,seq,marker,v->pcm_current,0,0,totalshift);
 300             }
 301
 302             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->stretchm[l]*.1;
 303             _analysis_output_always("stretch",seq,marker,v->pcm_current,0,0,totalshift);
 304
 305             seq++;
 306
 307           }
 308 #endif
 309
 310           ve->curmark=j;
 311           if(j>=testW)return(1);
 312           return(0);
 313         }
 314       }
 315       j+=ve->searchstep;
 316     }
 317   }
 318
 319   return(-1);
 320 }
 321
 322 int _ve_envelope_mark(vorbis_dsp_state *v){
 323   envelope_lookup *ve=((backend_lookup_state *)(v->backend_state))->ve;
 324   vorbis_info *vi=v->vi;
 325   codec_setup_info *ci=vi->codec_setup;
 326   long centerW=v->centerW;
 327   long beginW=centerW-ci->blocksizes[v->W]/4;
 328   long endW=centerW+ci->blocksizes[v->W]/4;
 329   if(v->W){
 330     beginW-=ci->blocksizes[v->lW]/4;
 331     endW+=ci->blocksizes[v->nW]/4;
 332   }else{
 333     beginW-=ci->blocksizes[0]/4;
 334     endW+=ci->blocksizes[0]/4;
 335   }
 336
 337   if(ve->curmark>=beginW && ve->curmark<endW)return(1);
 338   {
 339     long first=beginW/ve->searchstep;
 340     long last=endW/ve->searchstep;
 341     long i;
 342     for(i=first;i<last;i++)
 343       if(ve->mark[i])return(1);
 344   }
 345   return(0);
 346 }
 347
 348 void _ve_envelope_shift(envelope_lookup *e,long shift){
 349   int smallsize=e->current/e->searchstep;
 350   int smallshift=shift/e->searchstep;
 351   int i;
 352
 353   memmove(e->mark,e->mark+smallshift,(smallsize-smallshift)*sizeof(*e->mark));
 354
 355   #if 0
 356   for(i=0;i<VE_BANDS*e->ch;i++)
 357     memmove(e->filter[i].markers,
 358             e->filter[i].markers+smallshift,
 359             (1024-smallshift)*sizeof(*(*e->filter).markers));
 360   memmove(e->stretchm,
 361           e->stretchm+smallshift,
 362           (1024-smallshift)*sizeof(*e->stretchm));
 363   totalshift+=shift;
 364 #endif
 365
 366   e->current-=shift;
 367   if(e->curmark>=0)
 368     e->curmark-=shift;
 369   e->cursor-=shift;
 370 }
 371
 372
 373
 374
 375
 376