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 1.43 2002/03/23 03:17:33 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=fromdB(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(1300.f/22050.f*n/4.f)*2.f;
  57   e->band[0].end=rint(3000.f/22050.f*n/4.f)*2.f-e->band[0].begin;
  58   e->band[1].begin=rint(2600.f/22050.f*n/4.f)*2.f;
  59   e->band[1].end=rint(6000.f/22050.f*n/4.f)*2.f-e->band[1].begin;
  60   e->band[2].begin=rint(5300.f/22050.f*n/4.f)*2.f;
  61   e->band[2].end=rint(12000.f/22050.f*n/4.f)*2.f-e->band[2].begin;
  62   e->band[3].begin=rint(10600.f/22050.f*n/4.f)*2.f;
  63   e->band[3].end=rint(18000.f/22050.f*n/4.f)*2.f-e->band[3].begin;
  64
  65   e->band[0].window=_ogg_malloc((e->band[0].end)*sizeof(*e->band[0].window));
  66   e->band[1].window=_ogg_malloc((e->band[1].end)*sizeof(*e->band[1].window));
  67   e->band[2].window=_ogg_malloc((e->band[2].end)*sizeof(*e->band[2].window));
  68   e->band[3].window=_ogg_malloc((e->band[3].end)*sizeof(*e->band[3].window));
  69
  70   n=e->band[0].end;
  71   for(i=0;i<n;i++)
  72     e->band[0].window[i]=sin((i+.5)/n*M_PI);
  73   n=e->band[1].end;
  74   for(i=0;i<n;i++)
  75     e->band[1].window[i]=sin((i+.5)/n*M_PI);
  76   n=e->band[2].end;
  77   for(i=0;i<n;i++)
  78     e->band[2].window[i]=sin((i+.5)/n*M_PI);
  79   n=e->band[3].end;
  80   for(i=0;i<n;i++)
  81     e->band[3].window[i]=sin((i+.5)/n*M_PI);
  82
  83   e->filter=_ogg_calloc(VE_BANDS*ch,sizeof(*e->filter));
  84   e->mark=_ogg_calloc(e->storage,sizeof(*e->mark));
  85
  86
  87 }
  88
  89 void _ve_envelope_clear(envelope_lookup *e){
  90   int i;
  91   mdct_clear(&e->mdct);
  92   for(i=0;i<VE_BANDS;i++)
  93     _ogg_free(e->band[i].window);
  94   _ogg_free(e->mdct_win);
  95   _ogg_free(e->filter);
  96   _ogg_free(e->mark);
  97   memset(e,0,sizeof(*e));
  98 }
  99
 100 /* fairly straight threshhold-by-band based until we find something
 101    that works better and isn't patented. */
 102 static int seq2=0;
 103 static int _ve_amp(envelope_lookup *ve,
 104                    vorbis_info_psy_global *gi,
 105                    float *data,
 106                    envelope_band *bands,
 107                    envelope_filter_state *filters,
 108                    long pos){
 109   long n=ve->winlength;
 110   int ret=0;
 111   long i,j;
 112
 113   /* we want to have a 'minimum bar' for energy, else we're just
 114      basing blocks on quantization noise that outweighs the signal
 115      itself (for low power signals) */
 116
 117   float minV=ve->minenergy,acc[VE_BANDS];
 118   float *vec=alloca(n*sizeof(*vec));
 119   memset(acc,0,sizeof(acc));
 120
 121  /* window and transform */
 122   for(i=0;i<n;i++)
 123     vec[i]=data[i]*ve->mdct_win[i];
 124   mdct_forward(&ve->mdct,vec,vec);
 125
 126   /* accumulate amplitude by band */
 127   for(j=0;j<VE_BANDS;j++){
 128     for(i=0;i<bands[j].end;i++){
 129       float val=vec[i+bands[j].begin];
 130       acc[j]+=val*val*bands[j].window[i];
 131     }
 132     acc[j]/=i*.707f;
 133     if(acc[j]<minV*minV)acc[j]=minV*minV;
 134     acc[j]=todB(acc+j);
 135   }
 136
 137   /* convert amplitude to delta */
 138   for(j=0;j<VE_BANDS;j++){
 139     float val=acc[j]-filters[j].ampbuf[filters[j].ampptr];
 140     filters[j].ampbuf[filters[j].ampptr]=acc[j];
 141     acc[j]=val;
 142     filters[j].ampptr++;
 143     if(filters[j].ampptr>=VE_DIV)filters[j].ampptr=0;
 144   }
 145
 146   /* convolve deltas to threshhold values */
 147   for(j=0;j<VE_BANDS;j++){
 148     float *buf=filters[j].delbuf;
 149     float val=.14*buf[0]+.14*buf[1]+.72*acc[j];
 150     buf[0]=buf[1];buf[1]=acc[j];
 151     acc[j]=val;
 152     filters[j].markers[pos+1]=val;
 153   }
 154
 155   /* look at local min/max */
 156   for(j=0;j<VE_BANDS;j++){
 157     float *buf=filters[j].convbuf;
 158     if(buf[1]>gi->preecho_thresh[j] && buf[0]<buf[1] && acc[j]<buf[1])ret=1;
 159     if(buf[1]<gi->postecho_thresh[j] && buf[0]>buf[1] && acc[j]>buf[1])ret=1;
 160     buf[0]=buf[1];buf[1]=acc[j];
 161   }
 162   return(ret);
 163 }
 164
 165 static int seq=0;
 166 static ogg_int64_t totalshift=-1024;
 167
 168 long _ve_envelope_search(vorbis_dsp_state *v){
 169   vorbis_info *vi=v->vi;
 170   codec_setup_info *ci=vi->codec_setup;
 171   vorbis_info_psy_global *gi=&ci->psy_g_param;
 172   envelope_lookup *ve=((backend_lookup_state *)(v->backend_state))->ve;
 173   long i,j;
 174
 175   int first=ve->current/ve->searchstep;
 176   int last=v->pcm_current/ve->searchstep-VE_DIV;
 177   if(first<0)first=0;
 178
 179   /* make sure we have enough storage to match the PCM */
 180   if(last>ve->storage){
 181     ve->storage=last+VE_DIV;
 182     ve->mark=_ogg_realloc(ve->mark,ve->storage*sizeof(*ve->mark));
 183   }
 184
 185   for(j=first;j<last;j++){
 186     int ret=0;
 187     for(i=0;i<ve->ch;i++){
 188       /* the mark delay is one searchstep because of min/max finder */
 189       float *pcm=v->pcm[i]+ve->searchstep*(j+1);
 190       ret|=_ve_amp(ve,gi,pcm,ve->band,ve->filter+i*VE_BANDS,j);
 191     }
 192
 193     /* we assume a 'transient' occupies half a short block; this way,
 194        it's contained in two short blocks, else the first block is
 195        short and the second long, causing smearing */
 196     ve->mark[j+VE_DIV/2]=0;
 197     if(ret)
 198       for(i=0;i<=VE_DIV/2;i++)
 199         ve->mark[j+i]=ret;
 200   }
 201
 202   ve->current=last*ve->searchstep;
 203
 204   {
 205     long centerW=v->centerW;
 206     long testW=
 207       centerW+
 208       ci->blocksizes[v->W]/4+
 209       ci->blocksizes[1]/2+
 210       ci->blocksizes[0]/4;
 211
 212     j=ve->cursor;
 213
 214     while(j<ve->current){
 215       if(j>=testW)return(1);
 216       if(ve->mark[j/ve->searchstep]){
 217         if(j>centerW){
 218
 219 #if 0
 220           if(j>ve->curmark){
 221             float *marker=alloca(v->pcm_current*sizeof(*marker));
 222             int l;
 223             memset(marker,0,sizeof(*marker)*v->pcm_current);
 224
 225             fprintf(stderr,"mark! seq=%d, cursor:%fs time:%fs\n",
 226                     seq,
 227                     (totalshift+ve->cursor)/44100.,
 228                     (totalshift+j)/44100.);
 229
 230             _analysis_output_always("pcmL",seq,v->pcm[0],v->pcm_current,0,0,totalshift);
 231             _analysis_output_always("pcmR",seq,v->pcm[1],v->pcm_current,0,0,totalshift);
 232
 233             _analysis_output_always("markL",seq,v->pcm[0],j,0,0,totalshift);
 234             _analysis_output_always("markR",seq,v->pcm[1],j,0,0,totalshift);
 235
 236
 237             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[0].markers[l]*.01;
 238             _analysis_output_always("delL0",seq,marker,v->pcm_current,0,0,totalshift);
 239             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[1].markers[l]*.01;
 240             _analysis_output_always("delL1",seq,marker,v->pcm_current,0,0,totalshift);
 241             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[2].markers[l]*.01;
 242             _analysis_output_always("delL2",seq,marker,v->pcm_current,0,0,totalshift);
 243             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[3].markers[l]*.01;
 244             _analysis_output_always("delL3",seq,marker,v->pcm_current,0,0,totalshift);
 245             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[4].markers[l]*.01;
 246             _analysis_output_always("delR0",seq,marker,v->pcm_current,0,0,totalshift);
 247             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[5].markers[l]*.01;
 248             _analysis_output_always("delR1",seq,marker,v->pcm_current,0,0,totalshift);
 249             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[6].markers[l]*.01;
 250             _analysis_output_always("delR2",seq,marker,v->pcm_current,0,0,totalshift);
 251             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[7].markers[l]*.01;
 252             _analysis_output_always("delR3",seq,marker,v->pcm_current,0,0,totalshift);
 253
 254             seq++;
 255
 256           }
 257 #endif
 258
 259           ve->curmark=j;
 260           ve->cursor=j;
 261           if(j>=testW)return(1);
 262           return(0);
 263         }
 264       }
 265       j+=ve->searchstep;
 266     }
 267   }
 268   ve->cursor=j;
 269
 270   return(-1);
 271 }
 272
 273 int _ve_envelope_mark(vorbis_dsp_state *v){
 274   envelope_lookup *ve=((backend_lookup_state *)(v->backend_state))->ve;
 275   vorbis_info *vi=v->vi;
 276   codec_setup_info *ci=vi->codec_setup;
 277   long centerW=v->centerW;
 278   long beginW=centerW-ci->blocksizes[v->W]/4;
 279   long endW=centerW+ci->blocksizes[v->W]/4;
 280   if(v->W){
 281     beginW-=ci->blocksizes[v->lW]/4;
 282     endW+=ci->blocksizes[v->nW]/4;
 283   }else{
 284     beginW-=ci->blocksizes[0]/4;
 285     endW+=ci->blocksizes[0]/4;
 286   }
 287
 288   if(ve->curmark>=beginW && ve->curmark<endW)return(1);
 289   {
 290     long first=beginW/ve->searchstep;
 291     long last=endW/ve->searchstep;
 292     long i;
 293     for(i=first;i<last;i++)
 294       if(ve->mark[i])return(1);
 295   }
 296   return(0);
 297 }
 298
 299 void _ve_envelope_shift(envelope_lookup *e,long shift){
 300   int smallsize=e->current/e->searchstep+VE_DIV/2; /* VE_DIV/2 is to
 301                                                       match setting a
 302                                                       mark on a region
 303                                                       in
 304                                                       envelope_search */
 305   int smallshift=shift/e->searchstep;
 306   int i;
 307
 308   memmove(e->mark,e->mark+smallshift,(smallsize-smallshift)*sizeof(*e->mark));
 309
 310   for(i=0;i<VE_BANDS*e->ch;i++)
 311     memmove(e->filter[i].markers,
 312             e->filter[i].markers+smallshift,
 313             (1024-smallshift)*sizeof(*(*e->filter).markers));
 314
 315   e->current-=shift;
 316   if(e->curmark>=0)
 317     e->curmark-=shift;
 318   e->cursor-=shift;
 319   totalshift+=shift;
 320 }
 321
 322