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 1.47 2002/03/30 01:56:58 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=2;  e->band[0].end=4;
  54   e->band[1].begin=4;  e->band[1].end=8;
  55   e->band[2].begin=6;  e->band[2].end=10;
  56   e->band[3].begin=12; e->band[3].end=12;
  57   e->band[4].begin=18; e->band[4].end=16;
  58   e->band[5].begin=26; e->band[5].end=20;
  59   e->band[6].begin=36; e->band[6].end=24;
  60
  61   for(j=0;j<VE_BANDS;j++){
  62     n=e->band[j].end;
  63     e->band[j].window=_ogg_malloc(n*sizeof(*e->band[0].window));
  64     for(i=0;i<n;i++){
  65       e->band[j].window[i]=sin((i+.5)/n*M_PI);
  66       e->band[j].total+=e->band[j].window[i];
  67     }
  68     e->band[j].total=1./e->band[j].total;
  69   }
  70
  71   e->filter=_ogg_calloc(VE_BANDS*ch,sizeof(*e->filter));
  72   e->mark=_ogg_calloc(e->storage,sizeof(*e->mark));
  73
  74 }
  75
  76 void _ve_envelope_clear(envelope_lookup *e){
  77   int i;
  78   mdct_clear(&e->mdct);
  79   for(i=0;i<VE_BANDS;i++)
  80     _ogg_free(e->band[i].window);
  81   _ogg_free(e->mdct_win);
  82   _ogg_free(e->filter);
  83   _ogg_free(e->mark);
  84   memset(e,0,sizeof(*e));
  85 }
  86
  87 /* fairly straight threshhold-by-band based until we find something
  88    that works better and isn't patented. */
  89
  90 static int _ve_amp(envelope_lookup *ve,
  91                    vorbis_info_psy_global *gi,
  92                    float *data,
  93                    envelope_band *bands,
  94                    envelope_filter_state *filters,
  95                    long pos){
  96   long n=ve->winlength;
  97   int ret=0;
  98   long i,j;
  99   float decay;
 100
 101   /* we want to have a 'minimum bar' for energy, else we're just
 102      basing blocks on quantization noise that outweighs the signal
 103      itself (for low power signals) */
 104
 105   float minV=ve->minenergy;
 106   float *vec=alloca(n*sizeof(*vec));
 107
 108   /* stretch is used to gradually lengthen the number of windows
 109      considered prevoius-to-potential-trigger */
 110   int stretch=max(VE_MINSTRETCH,ve->stretch/2);
 111   float penalty=gi->stretch_penalty-(ve->stretch/2-VE_MINSTRETCH);
 112   if(penalty<0.f)penalty=0.f;
 113   if(penalty>gi->stretch_penalty)penalty=gi->stretch_penalty;
 114
 115   /*_analysis_output_always("lpcm",seq2,data,n,0,0,
 116     totalshift+pos*ve->searchstep);*/
 117
 118  /* window and transform */
 119   for(i=0;i<n;i++)
 120     vec[i]=data[i]*ve->mdct_win[i];
 121   mdct_forward(&ve->mdct,vec,vec);
 122
 123   /*_analysis_output_always("mdct",seq2,vec,n/2,0,1,0); */
 124
 125   /* near-DC spreading function; this has nothing to do with
 126      psychoacoustics, just sidelobe leakage and window size */
 127   {
 128     float temp=vec[0]*vec[0]+.7*vec[1]*vec[1]+.2*vec[2]*vec[2];
 129     int ptr=filters->nearptr;
 130
 131     /* the accumulation is regularly refreshed from scratch to avoid
 132        floating point creep */
 133     if(ptr==0){
 134       decay=filters->nearDC_acc=filters->nearDC_partialacc+temp;
 135       filters->nearDC_partialacc=temp;
 136     }else{
 137       decay=filters->nearDC_acc+=temp;
 138       filters->nearDC_partialacc+=temp;
 139     }
 140     filters->nearDC_acc-=filters->nearDC[ptr];
 141     filters->nearDC[ptr]=temp;
 142
 143     decay*=(1./(VE_NEARDC+1));
 144     filters->nearptr++;
 145     if(filters->nearptr>=VE_NEARDC)filters->nearptr=0;
 146     decay=todB(&decay)*.5-15.f;
 147   }
 148
 149   /* perform spreading and limiting, also smooth the spectrum.  yes,
 150      the MDCT results in all real coefficients, but it still *behaves*
 151      like real/imaginary pairs */
 152   for(i=0;i<n/2;i+=2){
 153     float val=vec[i]*vec[i]+vec[i+1]*vec[i+1];
 154     val=todB(&val)*.5f;
 155     if(val<decay)val=decay;
 156     if(val<minV)val=minV;
 157     vec[i>>1]=val;
 158     decay-=8.;
 159   }
 160
 161   /*_analysis_output_always("spread",seq2++,vec,n/4,0,0,0);*/
 162
 163   /* perform preecho/postecho triggering by band */
 164   for(j=0;j<VE_BANDS;j++){
 165     float acc=0.;
 166     float valmax,valmin;
 167
 168     /* accumulate amplitude */
 169     for(i=0;i<bands[j].end;i++)
 170       acc+=vec[i+bands[j].begin]*bands[j].window[i];
 171
 172     acc*=bands[j].total;
 173
 174     /* convert amplitude to delta */
 175     {
 176       int p,this=filters[j].ampptr;
 177       float postmax,postmin,premax=-99999.f,premin=99999.f;
 178
 179       p=this;
 180       p--;
 181       if(p<0)p+=VE_AMP;
 182       postmax=max(acc,filters[j].ampbuf[p]);
 183       postmin=min(acc,filters[j].ampbuf[p]);
 184
 185       for(i=0;i<stretch;i++){
 186         p--;
 187         if(p<0)p+=VE_AMP;
 188         premax=max(premax,filters[j].ampbuf[p]);
 189         premin=min(premin,filters[j].ampbuf[p]);
 190       }
 191
 192       valmin=postmin-premin;
 193       valmax=postmax-premax;
 194
 195       filters[j].markers[pos]=valmax;
 196
 197       filters[j].ampbuf[this]=acc;
 198       filters[j].ampptr++;
 199       if(filters[j].ampptr>=VE_AMP)filters[j].ampptr=0;
 200     }
 201
 202     /* look at min/max, decide trigger */
 203     if(valmax>gi->preecho_thresh[j]+penalty){
 204       ret|=1;
 205       ret|=4;
 206     }
 207     if(valmin<gi->postecho_thresh[j]-penalty)ret|=2;
 208   }
 209
 210   return(ret);
 211 }
 212
 213 long _ve_envelope_search(vorbis_dsp_state *v){
 214   vorbis_info *vi=v->vi;
 215   codec_setup_info *ci=vi->codec_setup;
 216   vorbis_info_psy_global *gi=&ci->psy_g_param;
 217   envelope_lookup *ve=((backend_lookup_state *)(v->backend_state))->ve;
 218   long i,j;
 219
 220   int first=ve->current/ve->searchstep;
 221   int last=v->pcm_current/ve->searchstep-VE_WIN;
 222   if(first<0)first=0;
 223
 224   /* make sure we have enough storage to match the PCM */
 225   if(last>ve->storage){
 226     ve->storage=last+VE_WIN;
 227     ve->mark=_ogg_realloc(ve->mark,ve->storage*sizeof(*ve->mark));
 228   }
 229
 230   for(j=first;j<last;j++){
 231     int ret=0;
 232
 233     ve->stretch++;
 234     if(ve->stretch>VE_MAXSTRETCH*2)
 235       ve->stretch=VE_MAXSTRETCH*2;
 236
 237     for(i=0;i<ve->ch;i++){
 238       float *pcm=v->pcm[i]+ve->searchstep*(j);
 239       ret|=_ve_amp(ve,gi,pcm,ve->band,ve->filter+i*VE_BANDS,j);
 240     }
 241
 242     ve->mark[j+VE_POST]=0;
 243     if(ret&1){
 244       ve->mark[j]=1;
 245       ve->mark[j+1]=1;
 246     }
 247
 248     if(ret&2){
 249       ve->mark[j]=1;
 250       if(j>0)ve->mark[j-1]=1;
 251     }
 252
 253     if(ret&4)ve->stretch=-1;
 254   }
 255
 256   ve->current=last*ve->searchstep;
 257
 258   {
 259     long centerW=v->centerW;
 260     long testW=
 261       centerW+
 262       ci->blocksizes[v->W]/4+
 263       ci->blocksizes[1]/2+
 264       ci->blocksizes[0]/4;
 265
 266     j=ve->cursor;
 267
 268     while(j<ve->current-(ve->searchstep)){/* account for postecho
 269                                              working back one window */
 270       if(j>=testW)return(1);
 271
 272       ve->cursor=j;
 273
 274       if(ve->mark[j/ve->searchstep]){
 275         if(j>centerW){
 276
 277 #if 0
 278           if(j>ve->curmark){
 279             float *marker=alloca(v->pcm_current*sizeof(*marker));
 280             int l,m;
 281             memset(marker,0,sizeof(*marker)*v->pcm_current);
 282             fprintf(stderr,"mark! seq=%d, cursor:%fs time:%fs\n",
 283                     seq,
 284                     (totalshift+ve->cursor)/44100.,
 285                     (totalshift+j)/44100.);
 286             _analysis_output_always("pcmL",seq,v->pcm[0],v->pcm_current,0,0,totalshift);
 287             _analysis_output_always("pcmR",seq,v->pcm[1],v->pcm_current,0,0,totalshift);
 288
 289             _analysis_output_always("markL",seq,v->pcm[0],j,0,0,totalshift);
 290             _analysis_output_always("markR",seq,v->pcm[1],j,0,0,totalshift);
 291
 292             for(m=0;m<VE_BANDS;m++){
 293               char buf[80];
 294               sprintf(buf,"delL%d",m);
 295               for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[m].markers[l]*.1;
 296               _analysis_output_always(buf,seq,marker,v->pcm_current,0,0,totalshift);
 297             }
 298
 299             for(m=0;m<VE_BANDS;m++){
 300               char buf[80];
 301               sprintf(buf,"delR%d",m);
 302               for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[m+VE_BANDS].markers[l]*.1;
 303               _analysis_output_always(buf,seq,marker,v->pcm_current,0,0,totalshift);
 304             }
 305
 306             seq++;
 307
 308           }
 309 #endif
 310
 311           ve->curmark=j;
 312           if(j>=testW)return(1);
 313           return(0);
 314         }
 315       }
 316       j+=ve->searchstep;
 317     }
 318   }
 319
 320   return(-1);
 321 }
 322
 323 int _ve_envelope_mark(vorbis_dsp_state *v){
 324   envelope_lookup *ve=((backend_lookup_state *)(v->backend_state))->ve;
 325   vorbis_info *vi=v->vi;
 326   codec_setup_info *ci=vi->codec_setup;
 327   long centerW=v->centerW;
 328   long beginW=centerW-ci->blocksizes[v->W]/4;
 329   long endW=centerW+ci->blocksizes[v->W]/4;
 330   if(v->W){
 331     beginW-=ci->blocksizes[v->lW]/4;
 332     endW+=ci->blocksizes[v->nW]/4;
 333   }else{
 334     beginW-=ci->blocksizes[0]/4;
 335     endW+=ci->blocksizes[0]/4;
 336   }
 337
 338   if(ve->curmark>=beginW && ve->curmark<endW)return(1);
 339   {
 340     long first=beginW/ve->searchstep;
 341     long last=endW/ve->searchstep;
 342     long i;
 343     for(i=first;i<last;i++)
 344       if(ve->mark[i])return(1);
 345   }
 346   return(0);
 347 }
 348
 349 void _ve_envelope_shift(envelope_lookup *e,long shift){
 350   int smallsize=e->current/e->searchstep;
 351   int smallshift=shift/e->searchstep;
 352   int i;
 353
 354   memmove(e->mark,e->mark+smallshift,(smallsize-smallshift)*sizeof(*e->mark));
 355
 356 #if 0
 357   for(i=0;i<VE_BANDS*e->ch;i++)
 358     memmove(e->filter[i].markers,
 359             e->filter[i].markers+smallshift,
 360             (1024-smallshift)*sizeof(*(*e->filter).markers));
 361   totalshift+=shift;
 362 #endif
 363
 364   e->current-=shift;
 365   if(e->curmark>=0)
 366     e->curmark-=shift;
 367   e->cursor-=shift;
 368 }
 369
 370
 371
 372
 373
 374