cannam@86: /********************************************************************
cannam@86:  *                                                                  *
cannam@86:  * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
cannam@86:  * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
cannam@86:  * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
cannam@86:  * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
cannam@86:  *                                                                  *
cannam@86:  * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2009             *
cannam@86:  * by the Xiph.Org Foundation http://www.xiph.org/                  *
cannam@86:  *                                                                  *
cannam@86:  ********************************************************************
cannam@86: 
cannam@86:  function: PCM data envelope analysis
cannam@86:  last mod: $Id: envelope.c 16227 2009-07-08 06:58:46Z xiphmont $
cannam@86: 
cannam@86:  ********************************************************************/
cannam@86: 
cannam@86: #include <stdlib.h>
cannam@86: #include <string.h>
cannam@86: #include <stdio.h>
cannam@86: #include <math.h>
cannam@86: #include <ogg/ogg.h>
cannam@86: #include "vorbis/codec.h"
cannam@86: #include "codec_internal.h"
cannam@86: 
cannam@86: #include "os.h"
cannam@86: #include "scales.h"
cannam@86: #include "envelope.h"
cannam@86: #include "mdct.h"
cannam@86: #include "misc.h"
cannam@86: 
cannam@86: void _ve_envelope_init(envelope_lookup *e,vorbis_info *vi){
cannam@86:   codec_setup_info *ci=vi->codec_setup;
cannam@86:   vorbis_info_psy_global *gi=&ci->psy_g_param;
cannam@86:   int ch=vi->channels;
cannam@86:   int i,j;
cannam@86:   int n=e->winlength=128;
cannam@86:   e->searchstep=64; /* not random */
cannam@86: 
cannam@86:   e->minenergy=gi->preecho_minenergy;
cannam@86:   e->ch=ch;
cannam@86:   e->storage=128;
cannam@86:   e->cursor=ci->blocksizes[1]/2;
cannam@86:   e->mdct_win=_ogg_calloc(n,sizeof(*e->mdct_win));
cannam@86:   mdct_init(&e->mdct,n);
cannam@86: 
cannam@86:   for(i=0;i<n;i++){
cannam@86:     e->mdct_win[i]=sin(i/(n-1.)*M_PI);
cannam@86:     e->mdct_win[i]*=e->mdct_win[i];
cannam@86:   }
cannam@86: 
cannam@86:   /* magic follows */
cannam@86:   e->band[0].begin=2;  e->band[0].end=4;
cannam@86:   e->band[1].begin=4;  e->band[1].end=5;
cannam@86:   e->band[2].begin=6;  e->band[2].end=6;
cannam@86:   e->band[3].begin=9;  e->band[3].end=8;
cannam@86:   e->band[4].begin=13;  e->band[4].end=8;
cannam@86:   e->band[5].begin=17;  e->band[5].end=8;
cannam@86:   e->band[6].begin=22;  e->band[6].end=8;
cannam@86: 
cannam@86:   for(j=0;j<VE_BANDS;j++){
cannam@86:     n=e->band[j].end;
cannam@86:     e->band[j].window=_ogg_malloc(n*sizeof(*e->band[0].window));
cannam@86:     for(i=0;i<n;i++){
cannam@86:       e->band[j].window[i]=sin((i+.5)/n*M_PI);
cannam@86:       e->band[j].total+=e->band[j].window[i];
cannam@86:     }
cannam@86:     e->band[j].total=1./e->band[j].total;
cannam@86:   }
cannam@86: 
cannam@86:   e->filter=_ogg_calloc(VE_BANDS*ch,sizeof(*e->filter));
cannam@86:   e->mark=_ogg_calloc(e->storage,sizeof(*e->mark));
cannam@86: 
cannam@86: }
cannam@86: 
cannam@86: void _ve_envelope_clear(envelope_lookup *e){
cannam@86:   int i;
cannam@86:   mdct_clear(&e->mdct);
cannam@86:   for(i=0;i<VE_BANDS;i++)
cannam@86:     _ogg_free(e->band[i].window);
cannam@86:   _ogg_free(e->mdct_win);
cannam@86:   _ogg_free(e->filter);
cannam@86:   _ogg_free(e->mark);
cannam@86:   memset(e,0,sizeof(*e));
cannam@86: }
cannam@86: 
cannam@86: /* fairly straight threshhold-by-band based until we find something
cannam@86:    that works better and isn't patented. */
cannam@86: 
cannam@86: static int _ve_amp(envelope_lookup *ve,
cannam@86:                    vorbis_info_psy_global *gi,
cannam@86:                    float *data,
cannam@86:                    envelope_band *bands,
cannam@86:                    envelope_filter_state *filters){
cannam@86:   long n=ve->winlength;
cannam@86:   int ret=0;
cannam@86:   long i,j;
cannam@86:   float decay;
cannam@86: 
cannam@86:   /* we want to have a 'minimum bar' for energy, else we're just
cannam@86:      basing blocks on quantization noise that outweighs the signal
cannam@86:      itself (for low power signals) */
cannam@86: 
cannam@86:   float minV=ve->minenergy;
cannam@86:   float *vec=alloca(n*sizeof(*vec));
cannam@86: 
cannam@86:   /* stretch is used to gradually lengthen the number of windows
cannam@86:      considered prevoius-to-potential-trigger */
cannam@86:   int stretch=max(VE_MINSTRETCH,ve->stretch/2);
cannam@86:   float penalty=gi->stretch_penalty-(ve->stretch/2-VE_MINSTRETCH);
cannam@86:   if(penalty<0.f)penalty=0.f;
cannam@86:   if(penalty>gi->stretch_penalty)penalty=gi->stretch_penalty;
cannam@86: 
cannam@86:   /*_analysis_output_always("lpcm",seq2,data,n,0,0,
cannam@86:     totalshift+pos*ve->searchstep);*/
cannam@86: 
cannam@86:  /* window and transform */
cannam@86:   for(i=0;i<n;i++)
cannam@86:     vec[i]=data[i]*ve->mdct_win[i];
cannam@86:   mdct_forward(&ve->mdct,vec,vec);
cannam@86: 
cannam@86:   /*_analysis_output_always("mdct",seq2,vec,n/2,0,1,0); */
cannam@86: 
cannam@86:   /* near-DC spreading function; this has nothing to do with
cannam@86:      psychoacoustics, just sidelobe leakage and window size */
cannam@86:   {
cannam@86:     float temp=vec[0]*vec[0]+.7*vec[1]*vec[1]+.2*vec[2]*vec[2];
cannam@86:     int ptr=filters->nearptr;
cannam@86: 
cannam@86:     /* the accumulation is regularly refreshed from scratch to avoid
cannam@86:        floating point creep */
cannam@86:     if(ptr==0){
cannam@86:       decay=filters->nearDC_acc=filters->nearDC_partialacc+temp;
cannam@86:       filters->nearDC_partialacc=temp;
cannam@86:     }else{
cannam@86:       decay=filters->nearDC_acc+=temp;
cannam@86:       filters->nearDC_partialacc+=temp;
cannam@86:     }
cannam@86:     filters->nearDC_acc-=filters->nearDC[ptr];
cannam@86:     filters->nearDC[ptr]=temp;
cannam@86: 
cannam@86:     decay*=(1./(VE_NEARDC+1));
cannam@86:     filters->nearptr++;
cannam@86:     if(filters->nearptr>=VE_NEARDC)filters->nearptr=0;
cannam@86:     decay=todB(&decay)*.5-15.f;
cannam@86:   }
cannam@86: 
cannam@86:   /* perform spreading and limiting, also smooth the spectrum.  yes,
cannam@86:      the MDCT results in all real coefficients, but it still *behaves*
cannam@86:      like real/imaginary pairs */
cannam@86:   for(i=0;i<n/2;i+=2){
cannam@86:     float val=vec[i]*vec[i]+vec[i+1]*vec[i+1];
cannam@86:     val=todB(&val)*.5f;
cannam@86:     if(val<decay)val=decay;
cannam@86:     if(val<minV)val=minV;
cannam@86:     vec[i>>1]=val;
cannam@86:     decay-=8.;
cannam@86:   }
cannam@86: 
cannam@86:   /*_analysis_output_always("spread",seq2++,vec,n/4,0,0,0);*/
cannam@86: 
cannam@86:   /* perform preecho/postecho triggering by band */
cannam@86:   for(j=0;j<VE_BANDS;j++){
cannam@86:     float acc=0.;
cannam@86:     float valmax,valmin;
cannam@86: 
cannam@86:     /* accumulate amplitude */
cannam@86:     for(i=0;i<bands[j].end;i++)
cannam@86:       acc+=vec[i+bands[j].begin]*bands[j].window[i];
cannam@86: 
cannam@86:     acc*=bands[j].total;
cannam@86: 
cannam@86:     /* convert amplitude to delta */
cannam@86:     {
cannam@86:       int p,this=filters[j].ampptr;
cannam@86:       float postmax,postmin,premax=-99999.f,premin=99999.f;
cannam@86: 
cannam@86:       p=this;
cannam@86:       p--;
cannam@86:       if(p<0)p+=VE_AMP;
cannam@86:       postmax=max(acc,filters[j].ampbuf[p]);
cannam@86:       postmin=min(acc,filters[j].ampbuf[p]);
cannam@86: 
cannam@86:       for(i=0;i<stretch;i++){
cannam@86:         p--;
cannam@86:         if(p<0)p+=VE_AMP;
cannam@86:         premax=max(premax,filters[j].ampbuf[p]);
cannam@86:         premin=min(premin,filters[j].ampbuf[p]);
cannam@86:       }
cannam@86: 
cannam@86:       valmin=postmin-premin;
cannam@86:       valmax=postmax-premax;
cannam@86: 
cannam@86:       /*filters[j].markers[pos]=valmax;*/
cannam@86:       filters[j].ampbuf[this]=acc;
cannam@86:       filters[j].ampptr++;
cannam@86:       if(filters[j].ampptr>=VE_AMP)filters[j].ampptr=0;
cannam@86:     }
cannam@86: 
cannam@86:     /* look at min/max, decide trigger */
cannam@86:     if(valmax>gi->preecho_thresh[j]+penalty){
cannam@86:       ret|=1;
cannam@86:       ret|=4;
cannam@86:     }
cannam@86:     if(valmin<gi->postecho_thresh[j]-penalty)ret|=2;
cannam@86:   }
cannam@86: 
cannam@86:   return(ret);
cannam@86: }
cannam@86: 
cannam@86: #if 0
cannam@86: static int seq=0;
cannam@86: static ogg_int64_t totalshift=-1024;
cannam@86: #endif
cannam@86: 
cannam@86: long _ve_envelope_search(vorbis_dsp_state *v){
cannam@86:   vorbis_info *vi=v->vi;
cannam@86:   codec_setup_info *ci=vi->codec_setup;
cannam@86:   vorbis_info_psy_global *gi=&ci->psy_g_param;
cannam@86:   envelope_lookup *ve=((private_state *)(v->backend_state))->ve;
cannam@86:   long i,j;
cannam@86: 
cannam@86:   int first=ve->current/ve->searchstep;
cannam@86:   int last=v->pcm_current/ve->searchstep-VE_WIN;
cannam@86:   if(first<0)first=0;
cannam@86: 
cannam@86:   /* make sure we have enough storage to match the PCM */
cannam@86:   if(last+VE_WIN+VE_POST>ve->storage){
cannam@86:     ve->storage=last+VE_WIN+VE_POST; /* be sure */
cannam@86:     ve->mark=_ogg_realloc(ve->mark,ve->storage*sizeof(*ve->mark));
cannam@86:   }
cannam@86: 
cannam@86:   for(j=first;j<last;j++){
cannam@86:     int ret=0;
cannam@86: 
cannam@86:     ve->stretch++;
cannam@86:     if(ve->stretch>VE_MAXSTRETCH*2)
cannam@86:       ve->stretch=VE_MAXSTRETCH*2;
cannam@86: 
cannam@86:     for(i=0;i<ve->ch;i++){
cannam@86:       float *pcm=v->pcm[i]+ve->searchstep*(j);
cannam@86:       ret|=_ve_amp(ve,gi,pcm,ve->band,ve->filter+i*VE_BANDS);
cannam@86:     }
cannam@86: 
cannam@86:     ve->mark[j+VE_POST]=0;
cannam@86:     if(ret&1){
cannam@86:       ve->mark[j]=1;
cannam@86:       ve->mark[j+1]=1;
cannam@86:     }
cannam@86: 
cannam@86:     if(ret&2){
cannam@86:       ve->mark[j]=1;
cannam@86:       if(j>0)ve->mark[j-1]=1;
cannam@86:     }
cannam@86: 
cannam@86:     if(ret&4)ve->stretch=-1;
cannam@86:   }
cannam@86: 
cannam@86:   ve->current=last*ve->searchstep;
cannam@86: 
cannam@86:   {
cannam@86:     long centerW=v->centerW;
cannam@86:     long testW=
cannam@86:       centerW+
cannam@86:       ci->blocksizes[v->W]/4+
cannam@86:       ci->blocksizes[1]/2+
cannam@86:       ci->blocksizes[0]/4;
cannam@86: 
cannam@86:     j=ve->cursor;
cannam@86: 
cannam@86:     while(j<ve->current-(ve->searchstep)){/* account for postecho
cannam@86:                                              working back one window */
cannam@86:       if(j>=testW)return(1);
cannam@86: 
cannam@86:       ve->cursor=j;
cannam@86: 
cannam@86:       if(ve->mark[j/ve->searchstep]){
cannam@86:         if(j>centerW){
cannam@86: 
cannam@86: #if 0
cannam@86:           if(j>ve->curmark){
cannam@86:             float *marker=alloca(v->pcm_current*sizeof(*marker));
cannam@86:             int l,m;
cannam@86:             memset(marker,0,sizeof(*marker)*v->pcm_current);
cannam@86:             fprintf(stderr,"mark! seq=%d, cursor:%fs time:%fs\n",
cannam@86:                     seq,
cannam@86:                     (totalshift+ve->cursor)/44100.,
cannam@86:                     (totalshift+j)/44100.);
cannam@86:             _analysis_output_always("pcmL",seq,v->pcm[0],v->pcm_current,0,0,totalshift);
cannam@86:             _analysis_output_always("pcmR",seq,v->pcm[1],v->pcm_current,0,0,totalshift);
cannam@86: 
cannam@86:             _analysis_output_always("markL",seq,v->pcm[0],j,0,0,totalshift);
cannam@86:             _analysis_output_always("markR",seq,v->pcm[1],j,0,0,totalshift);
cannam@86: 
cannam@86:             for(m=0;m<VE_BANDS;m++){
cannam@86:               char buf[80];
cannam@86:               sprintf(buf,"delL%d",m);
cannam@86:               for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[m].markers[l]*.1;
cannam@86:               _analysis_output_always(buf,seq,marker,v->pcm_current,0,0,totalshift);
cannam@86:             }
cannam@86: 
cannam@86:             for(m=0;m<VE_BANDS;m++){
cannam@86:               char buf[80];
cannam@86:               sprintf(buf,"delR%d",m);
cannam@86:               for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->filter[m+VE_BANDS].markers[l]*.1;
cannam@86:               _analysis_output_always(buf,seq,marker,v->pcm_current,0,0,totalshift);
cannam@86:             }
cannam@86: 
cannam@86:             for(l=0;l<last;l++)marker[l*ve->searchstep]=ve->mark[l]*.4;
cannam@86:             _analysis_output_always("mark",seq,marker,v->pcm_current,0,0,totalshift);
cannam@86: 
cannam@86: 
cannam@86:             seq++;
cannam@86: 
cannam@86:           }
cannam@86: #endif
cannam@86: 
cannam@86:           ve->curmark=j;
cannam@86:           if(j>=testW)return(1);
cannam@86:           return(0);
cannam@86:         }
cannam@86:       }
cannam@86:       j+=ve->searchstep;
cannam@86:     }
cannam@86:   }
cannam@86: 
cannam@86:   return(-1);
cannam@86: }
cannam@86: 
cannam@86: int _ve_envelope_mark(vorbis_dsp_state *v){
cannam@86:   envelope_lookup *ve=((private_state *)(v->backend_state))->ve;
cannam@86:   vorbis_info *vi=v->vi;
cannam@86:   codec_setup_info *ci=vi->codec_setup;
cannam@86:   long centerW=v->centerW;
cannam@86:   long beginW=centerW-ci->blocksizes[v->W]/4;
cannam@86:   long endW=centerW+ci->blocksizes[v->W]/4;
cannam@86:   if(v->W){
cannam@86:     beginW-=ci->blocksizes[v->lW]/4;
cannam@86:     endW+=ci->blocksizes[v->nW]/4;
cannam@86:   }else{
cannam@86:     beginW-=ci->blocksizes[0]/4;
cannam@86:     endW+=ci->blocksizes[0]/4;
cannam@86:   }
cannam@86: 
cannam@86:   if(ve->curmark>=beginW && ve->curmark<endW)return(1);
cannam@86:   {
cannam@86:     long first=beginW/ve->searchstep;
cannam@86:     long last=endW/ve->searchstep;
cannam@86:     long i;
cannam@86:     for(i=first;i<last;i++)
cannam@86:       if(ve->mark[i])return(1);
cannam@86:   }
cannam@86:   return(0);
cannam@86: }
cannam@86: 
cannam@86: void _ve_envelope_shift(envelope_lookup *e,long shift){
cannam@86:   int smallsize=e->current/e->searchstep+VE_POST; /* adjust for placing marks
cannam@86:                                                      ahead of ve->current */
cannam@86:   int smallshift=shift/e->searchstep;
cannam@86: 
cannam@86:   memmove(e->mark,e->mark+smallshift,(smallsize-smallshift)*sizeof(*e->mark));
cannam@86: 
cannam@86: #if 0
cannam@86:   for(i=0;i<VE_BANDS*e->ch;i++)
cannam@86:     memmove(e->filter[i].markers,
cannam@86:             e->filter[i].markers+smallshift,
cannam@86:             (1024-smallshift)*sizeof(*(*e->filter).markers));
cannam@86:   totalshift+=shift;
cannam@86: #endif
cannam@86: 
cannam@86:   e->current-=shift;
cannam@86:   if(e->curmark>=0)
cannam@86:     e->curmark-=shift;
cannam@86:   e->cursor-=shift;
cannam@86: }