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 1.44 2002/03/24 21:04:00 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;
  37   int n=e->winlength=ci->blocksizes[0];
  38   e->searchstep=ci->blocksizes[0]/VE_DIV; /* 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+.5)/n*M_PI);
  49     e->mdct_win[i]*=e->mdct_win[i];
  50   }
  51
  52   /* overlapping bands, assuming 22050 (which is not always true, but
  53      to Hell with that) */
  54   /* 2(1.3-3) 4(2.6-6) 8(5.3-12) 16(10.6-18) */
  55
  56   e->band[0].begin=rint(3000.f/22050.f*n/4.f)*2.f;
  57   e->band[0].end=rint(6000.f/22050.f*n/4.f)*2.f-e->band[0].begin;
  58   e->band[1].begin=rint(5000.f/22050.f*n/4.f)*2.f;
  59   e->band[1].end=rint(10000.f/22050.f*n/4.f)*2.f-e->band[1].begin;
  60   e->band[2].begin=rint(8000.f/22050.f*n/4.f)*2.f;
  61   e->band[2].end=rint(16000.f/22050.f*n/4.f)*2.f-e->band[2].begin;
  62   e->band[3].begin=rint(12000.f/22050.f*n/4.f)*2.f;
  63   e->band[3].end=rint(20000.f/22050.f*n/4.f)*2.f-e->band[3].begin;
  64
  65   e->band[0].window=_ogg_malloc((e->band[0].end)/2*sizeof(*e->band[0].window));
  66   e->band[1].window=_ogg_malloc((e->band[1].end)/2*sizeof(*e->band[1].window));
  67   e->band[2].window=_ogg_malloc((e->band[2].end)/2*sizeof(*e->band[2].window));
  68   e->band[3].window=_ogg_malloc((e->band[3].end)/2*sizeof(*e->band[3].window));
  69
  70   n=e->band[0].end/2;
  71   for(i=0;i<n;i++){
  72     e->band[0].window[i]=sin((i+.5)/n*M_PI);
  73     e->band[0].total+=e->band[0].window[i];
  74   }
  75   n=e->band[1].end/2;
  76   for(i=0;i<n;i++){
  77     e->band[1].window[i]=sin((i+.5)/n*M_PI);
  78     e->band[1].total+=e->band[1].window[i];
  79   }
  80   n=e->band[2].end/2;
  81   for(i=0;i<n;i++){
  82     e->band[2].window[i]=sin((i+.5)/n*M_PI);
  83     e->band[2].total+=e->band[2].window[i];
  84   }
  85   n=e->band[3].end/2;
  86   for(i=0;i<n;i++){
  87     e->band[3].window[i]=sin((i+.5)/n*M_PI);
  88     e->band[3].total+=e->band[3].window[i];
  89   }
  90
  91
  92   e->filter=_ogg_calloc(VE_BANDS*ch,sizeof(*e->filter));
  93   e->mark=_ogg_calloc(e->storage,sizeof(*e->mark));
  94
  95
  96 }
  97
  98 void _ve_envelope_clear(envelope_lookup *e){
  99   int i;
 100   mdct_clear(&e->mdct);
 101   for(i=0;i<VE_BANDS;i++)
 102     _ogg_free(e->band[i].window);
 103   _ogg_free(e->mdct_win);
 104   _ogg_free(e->filter);
 105   _ogg_free(e->mark);
 106   memset(e,0,sizeof(*e));
 107 }
 108
 109 /* fairly straight threshhold-by-band based until we find something
 110    that works better and isn't patented. */
 111 static int _ve_amp(envelope_lookup *ve,
 112                    vorbis_info_psy_global *gi,
 113                    float *data,
 114                    envelope_band *bands,
 115                    envelope_filter_state *filters,
 116                    long pos){
 117   long n=ve->winlength;
 118   int ret=0;
 119   long i,j;
 120
 121   /* we want to have a 'minimum bar' for energy, else we're just
 122      basing blocks on quantization noise that outweighs the signal
 123      itself (for low power signals) */
 124
 125   float minV=ve->minenergy,acc[VE_BANDS];
 126   float *vec=alloca(n*sizeof(*vec));
 127   memset(acc,0,sizeof(acc));
 128
 129  /* window and transform */
 130   for(i=0;i<n;i++)
 131     vec[i]=data[i]*ve->mdct_win[i];
 132   mdct_forward(&ve->mdct,vec,vec);
 133
 134   /* accumulate amplitude by band */
 135   for(j=0;j<VE_BANDS;j++){
 136     for(i=0;i<bands[j].end;i+=2){
 137       float val=FABS(vec+i+bands[j].begin)+FABS(vec+i+bands[j].begin+1);
 138       acc[j]+=todB(&val)*bands[j].window[i>>1];
 139     }
 140     acc[j]/=bands[j].total;
 141     if(acc[j]<minV)acc[j]=minV;
 142     //fprintf(stderr,"%d %gdB :: ",j,acc[j]);
 143   }
 144   //fprintf(stderr,"\n");
 145
 146   /* convert amplitude to delta */
 147   for(j=0;j<VE_BANDS;j++){
 148     float val=acc[j]-filters[j].ampbuf[filters[j].ampptr];
 149     filters[j].ampbuf[filters[j].ampptr]=acc[j];
 150     acc[j]=val;
 151     filters[j].ampptr++;
 152     if(filters[j].ampptr>=VE_DIV)filters[j].ampptr=0;
 153   }
 154
 155   /* convolve deltas to threshhold values */
 156   for(j=0;j<VE_BANDS;j++){
 157     float *buf=filters[j].delbuf;
 158     float val=.14*buf[0]+.14*buf[1]+.72*acc[j];
 159     buf[0]=buf[1];buf[1]=acc[j];
 160     acc[j]=val;
 161     /*filters[j].markers[pos+1]=val;*/
 162   }
 163
 164   /* look at local min/max */
 165   for(j=0;j<VE_BANDS;j++){
 166     float *buf=filters[j].convbuf;
 167     if(buf[1]>gi->preecho_thresh[j] && buf[0]<buf[1] && acc[j]<buf[1])ret|=1;
 168     if(buf[1]<gi->postecho_thresh[j] && buf[0]>buf[1] && acc[j]>buf[1])ret|=2;
 169     buf[0]=buf[1];buf[1]=acc[j];
 170   }
 171   return(ret);
 172 }
 173
 174 long _ve_envelope_search(vorbis_dsp_state *v){
 175   vorbis_info *vi=v->vi;
 176   codec_setup_info *ci=vi->codec_setup;
 177   vorbis_info_psy_global *gi=&ci->psy_g_param;
 178   envelope_lookup *ve=((backend_lookup_state *)(v->backend_state))->ve;
 179   long i,j;
 180
 181   int first=ve->current/ve->searchstep;
 182   int last=v->pcm_current/ve->searchstep-VE_DIV;
 183   if(first<0)first=0;
 184
 185   /* make sure we have enough storage to match the PCM */
 186   if(last>ve->storage){
 187     ve->storage=last+VE_DIV;
 188     ve->mark=_ogg_realloc(ve->mark,ve->storage*sizeof(*ve->mark));
 189   }
 190
 191   for(j=first;j<last;j++){
 192     int ret=0;
 193     for(i=0;i<ve->ch;i++){
 194       /* the mark delay is one searchstep because of min/max finder */
 195       float *pcm=v->pcm[i]+ve->searchstep*(j+1);
 196       ret|=_ve_amp(ve,gi,pcm,ve->band,ve->filter+i*VE_BANDS,j);
 197     }
 198
 199     /* we assume a 'transient' occupies half a short block; this way,
 200        it's contained in two short blocks, else the first block is
 201        short and the second long, causing smearing.
 202
 203       preecho triggers follow the impulse marker; postecho triger preceed it */
 204
 205     ve->mark[j+VE_DIV/2]=0;
 206     if(ret&1)
 207       for(i=0;i<=VE_DIV/2;i++)
 208         ve->mark[j+i]=1;
 209     if(ret&2)
 210       for(i=-1;i>=-VE_DIV/2 && j+i>=0;i--)
 211         ve->mark[j+i]=1;
 212   }
 213
 214   ve->current=last*ve->searchstep;
 215
 216   {
 217     long centerW=v->centerW;
 218     long testW=
 219       centerW+
 220       ci->blocksizes[v->W]/4+
 221       ci->blocksizes[1]/2+
 222       ci->blocksizes[0]/4;
 223
 224     j=ve->cursor;
 225
 226     while(j<ve->current-(VE_DIV/2*ve->searchstep)){ /* modified to
 227                                                        stay clear of
 228                                                        possibly
 229                                                        unfinished
 230                                                        postecho
 231                                                        detection */
 232       if(j>=testW)return(1);
 233       if(ve->mark[j/ve->searchstep]){
 234         if(j>centerW){
 235
 236 #if 0
 237           if(j>ve->curmark){
 238             float *marker=alloca(v->pcm_current*sizeof(*marker));
 239             int l;
 240             memset(marker,0,sizeof(*marker)*v->pcm_current);
 241             fprintf(stderr,"mark! seq=%d, cursor:%fs time:%fs\n",
 242                     seq,
 243                     (totalshift+ve->cursor)/44100.,
 244                     (totalshift+j)/44100.);
 245             _analysis_output_always("pcmL",seq,v->pcm[0],v->pcm_current,0,0,totalshift);
 246             _analysis_output_always("pcmR",seq,v->pcm[1],v->pcm_current,0,0,totalshift);
 247
 248             _analysis_output_always("markL",seq,v->pcm[0],j,0,0,totalshift);
 249             _analysis_output_always("markR",seq,v->pcm[1],j,0,0,totalshift);
 250
 251
 252             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[0].markers[l]*.01;
 253             _analysis_output_always("delL0",seq,marker,v->pcm_current,0,0,totalshift);
 254             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[1].markers[l]*.01;
 255             _analysis_output_always("delL1",seq,marker,v->pcm_current,0,0,totalshift);
 256             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[2].markers[l]*.01;
 257             _analysis_output_always("delL2",seq,marker,v->pcm_current,0,0,totalshift);
 258             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[3].markers[l]*.01;
 259             _analysis_output_always("delL3",seq,marker,v->pcm_current,0,0,totalshift);
 260             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[4].markers[l]*.01;
 261             _analysis_output_always("delR0",seq,marker,v->pcm_current,0,0,totalshift);
 262             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[5].markers[l]*.01;
 263             _analysis_output_always("delR1",seq,marker,v->pcm_current,0,0,totalshift);
 264             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[6].markers[l]*.01;
 265             _analysis_output_always("delR2",seq,marker,v->pcm_current,0,0,totalshift);
 266             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[7].markers[l]*.01;
 267             _analysis_output_always("delR3",seq,marker,v->pcm_current,0,0,totalshift);
 268             seq++;
 269
 270           }
 271 #endif
 272
 273           ve->curmark=j;
 274           ve->cursor=j;
 275           if(j>=testW)return(1);
 276           return(0);
 277         }
 278       }
 279       j+=ve->searchstep;
 280     }
 281   }
 282   ve->cursor=j;
 283
 284   return(-1);
 285 }
 286
 287 int _ve_envelope_mark(vorbis_dsp_state *v){
 288   envelope_lookup *ve=((backend_lookup_state *)(v->backend_state))->ve;
 289   vorbis_info *vi=v->vi;
 290   codec_setup_info *ci=vi->codec_setup;
 291   long centerW=v->centerW;
 292   long beginW=centerW-ci->blocksizes[v->W]/4;
 293   long endW=centerW+ci->blocksizes[v->W]/4;
 294   if(v->W){
 295     beginW-=ci->blocksizes[v->lW]/4;
 296     endW+=ci->blocksizes[v->nW]/4;
 297   }else{
 298     beginW-=ci->blocksizes[0]/4;
 299     endW+=ci->blocksizes[0]/4;
 300   }
 301
 302   if(ve->curmark>=beginW && ve->curmark<endW)return(1);
 303   {
 304     long first=beginW/ve->searchstep;
 305     long last=endW/ve->searchstep;
 306     long i;
 307     for(i=first;i<last;i++)
 308       if(ve->mark[i])return(1);
 309   }
 310   return(0);
 311 }
 312
 313 void _ve_envelope_shift(envelope_lookup *e,long shift){
 314   int smallsize=e->current/e->searchstep+VE_DIV/2; /* VE_DIV/2 is to
 315                                                       match setting a
 316                                                       mark on a region
 317                                                       in
 318                                                       envelope_search */
 319   int smallshift=shift/e->searchstep;
 320   int i;
 321
 322   memmove(e->mark,e->mark+smallshift,(smallsize-smallshift)*sizeof(*e->mark));
 323
 324 #if 0
 325   for(i=0;i<VE_BANDS*e->ch;i++)
 326     memmove(e->filter[i].markers,
 327             e->filter[i].markers+smallshift,
 328             (1024-smallshift)*sizeof(*(*e->filter).markers));
 329   totalshift+=shift;
 330 #endif
 331
 332   e->current-=shift;
 333   if(e->curmark>=0)
 334     e->curmark-=shift;
 335   e->cursor-=shift;
 336 }
 337
 338