nb_celp.c

这是著名的TCPMP播放器在WINDWOWS,和WINCE下编译通过的源程序.笔者对其中的LIBMAD库做了针对ARM MPU的优化. 并增加了词幕功能.

            pit_max=offset;

#ifdef EPIC_48K
         if (st->lbr_48k)
         {
            pitch = SUBMODE(ltp_quant)(target, sw, st->interp_qlpc, st->bw_lpc1, st->bw_lpc2,
                                       exc, SUBMODE(ltp_params), pit_min, pit_max, ol_pitch_coef,
                                       st->lpcSize, st->subframeSize, bits, stack, 
                                       exc, syn_resp, st->complexity, ol_pitch_id, st->plc_tuning);
         } else {
#endif

         /* Perform pitch search */
         pitch = SUBMODE(ltp_quant)(target, sw, st->interp_qlpc, st->bw_lpc1, st->bw_lpc2,
                                    exc, SUBMODE(ltp_params), pit_min, pit_max, ol_pitch_coef,
                                    st->lpcSize, st->subframeSize, bits, stack, 
                                    exc, syn_resp, st->complexity, 0, st->plc_tuning);
#ifdef EPIC_48K
         }
#endif

         st->pitch[sub]=pitch;
      } else {
         speex_error ("No pitch prediction, what's wrong");
      }

      /* Quantization of innovation */
      {
         spx_sig_t *innov;
         spx_word32_t ener=0;
         spx_word16_t fine_gain;

         innov = st->innov+sub*st->subframeSize;
         for (i=0;i<st->subframeSize;i++)
            innov[i]=0;
         
         for (i=0;i<st->subframeSize;i++)
            real_exc[i] = SUB32(real_exc[i], exc[i]);

         ener = SHL32(EXTEND32(compute_rms(real_exc, st->subframeSize)),SIG_SHIFT);
         
         /*FIXME: Should use DIV32_16 and make sure result fits in 16 bits */
#ifdef FIXED_POINT
         {
            spx_word32_t f = DIV32(ener,PSHR32(ol_gain,SIG_SHIFT));
            if (f<=32767)
               fine_gain = f;
            else
               fine_gain = 32767;
         }
#else
         fine_gain = DIV32_16(ener,PSHR32(ol_gain,SIG_SHIFT));
#endif
         /* Calculate gain correction for the sub-frame (if any) */
         if (SUBMODE(have_subframe_gain)) 
         {
            int qe;
            if (SUBMODE(have_subframe_gain)==3)
            {
               qe = scal_quant(fine_gain, exc_gain_quant_scal3_bound, 8);
               speex_bits_pack(bits, qe, 3);
               ener=MULT16_32_Q14(exc_gain_quant_scal3[qe],ol_gain);
            } else {
               qe = scal_quant(fine_gain, exc_gain_quant_scal1_bound, 2);
               speex_bits_pack(bits, qe, 1);
               ener=MULT16_32_Q14(exc_gain_quant_scal1[qe],ol_gain);               
            }
         } else {
            ener=ol_gain;
         }

         /*printf ("%f %f\n", ener, ol_gain);*/

         /* Normalize innovation */
         signal_div(target, target, ener, st->subframeSize);

         /* Quantize innovation */
         if (SUBMODE(innovation_quant))
         {
            /* Codebook search */
            SUBMODE(innovation_quant)(target, st->interp_qlpc, st->bw_lpc1, st->bw_lpc2, 
                                      SUBMODE(innovation_params), st->lpcSize, st->subframeSize, 
                                      innov, syn_resp, bits, stack, st->complexity, SUBMODE(double_codebook));
            
            /* De-normalize innovation and update excitation */
            signal_mul(innov, innov, ener, st->subframeSize);

            for (i=0;i<st->subframeSize;i++)
               exc[i] = ADD32(exc[i],innov[i]);
         } else {
            speex_error("No fixed codebook");
         }

         /* In some (rare) modes, we do a second search (more bits) to reduce noise even more */
         if (SUBMODE(double_codebook)) {
            char *tmp_stack=stack;
            VARDECL(spx_sig_t *innov2);
            ALLOC(innov2, st->subframeSize, spx_sig_t);
            for (i=0;i<st->subframeSize;i++)
               innov2[i]=0;
            for (i=0;i<st->subframeSize;i++)
               target[i]*=2.2;
            SUBMODE(innovation_quant)(target, st->interp_qlpc, st->bw_lpc1, st->bw_lpc2, 
                                      SUBMODE(innovation_params), st->lpcSize, st->subframeSize, 
                                      innov2, syn_resp, bits, stack, st->complexity, 0);
            signal_mul(innov2, innov2, (spx_word32_t) (ener*(1/2.2)), st->subframeSize);
            for (i=0;i<st->subframeSize;i++)
               exc[i] = ADD32(exc[i],innov2[i]);
            stack = tmp_stack;
         }

      }

      /* Final signal synthesis from excitation */
      iir_mem2(exc, st->interp_qlpc, sp, st->subframeSize, st->lpcSize, st->mem_sp);

      /* Compute weighted signal again, from synthesized speech (not sure it's the right thing) */
      if (st->complexity!=0)
         filter_mem2(sp, st->bw_lpc1, st->bw_lpc2, sw, st->subframeSize, st->lpcSize, st->mem_sw);
      
   }

   /* Store the LSPs for interpolation in the next frame */
   if (st->submodeID>=1)
   {
      for (i=0;i<st->lpcSize;i++)
         st->old_lsp[i] = st->lsp[i];
      for (i=0;i<st->lpcSize;i++)
         st->old_qlsp[i] = st->qlsp[i];
   }

   if (st->submodeID==1)
   {
      if (st->dtx_count)
         speex_bits_pack(bits, 15, 4);
      else
         speex_bits_pack(bits, 0, 4);
   }

   /* The next frame will not be the first (Duh!) */
   st->first = 0;

#ifdef RESYNTH
   /* Replace input by synthesized speech */
   for (i=0;i<st->frameSize;i++)
   {
      spx_word32_t sig = PSHR32(st->frame[i],SIG_SHIFT);
      if (sig>32767)
         sig = 32767;
      if (sig<-32767)
         sig = -32767;
     in[i]=sig;
   }
#endif

   if (SUBMODE(innovation_quant) == noise_codebook_quant || st->submodeID==0)
      st->bounded_pitch = 1;
   else
      st->bounded_pitch = 0;

   return 1;
}


void *nb_decoder_init(const SpeexMode *m)
{
   DecState *st;
   const SpeexNBMode *mode;
   int i;

   mode=(const SpeexNBMode*)m->mode;
   st = (DecState *)speex_alloc(sizeof(DecState));
   if (!st)
      return NULL;
#if defined(VAR_ARRAYS) || defined (USE_ALLOCA)
   st->stack = NULL;
#else
   st->stack = (char*)speex_alloc_scratch(NB_DEC_STACK);
#endif

   st->mode=m;


   st->encode_submode = 1;
#ifdef EPIC_48K
   st->lbr_48k=mode->lbr48k;
#endif

   st->first=1;
   /* Codec parameters, should eventually have several "modes"*/
   st->frameSize = mode->frameSize;
   st->nbSubframes=mode->frameSize/mode->subframeSize;
   st->subframeSize=mode->subframeSize;
   st->lpcSize = mode->lpcSize;
   st->min_pitch=mode->pitchStart;
   st->max_pitch=mode->pitchEnd;

   st->submodes=mode->submodes;
   st->submodeID=mode->defaultSubmode;

   st->lpc_enh_enabled=0;


   st->inBuf = speex_alloc((st->frameSize)*sizeof(spx_sig_t));
   st->frame = st->inBuf;
   st->excBuf = speex_alloc((st->frameSize + st->max_pitch + 1)*sizeof(spx_sig_t));
   st->exc = st->excBuf + st->max_pitch + 1;
   for (i=0;i<st->frameSize;i++)
      st->inBuf[i]=0;
   for (i=0;i<st->frameSize + st->max_pitch + 1;i++)
      st->excBuf[i]=0;
   st->innov = speex_alloc((st->frameSize)*sizeof(spx_sig_t));

   st->interp_qlpc = speex_alloc(st->lpcSize*sizeof(spx_coef_t));
   st->qlsp = speex_alloc(st->lpcSize*sizeof(spx_lsp_t));
   st->old_qlsp = speex_alloc(st->lpcSize*sizeof(spx_lsp_t));
   st->interp_qlsp = speex_alloc(st->lpcSize*sizeof(spx_lsp_t));
   st->mem_sp = speex_alloc((5*st->lpcSize)*sizeof(spx_mem_t));
   st->comb_mem = speex_alloc(sizeof(CombFilterMem));
   comb_filter_mem_init (st->comb_mem);

   st->pi_gain = speex_alloc((st->nbSubframes)*sizeof(spx_word32_t));
   st->last_pitch = 40;
   st->count_lost=0;
   st->pitch_gain_buf[0] = st->pitch_gain_buf[1] = st->pitch_gain_buf[2] = 0;
   st->pitch_gain_buf_idx = 0;
   st->seed = 1000;
   
   st->sampling_rate=8000;
   st->last_ol_gain = 0;

   st->user_callback.func = &speex_default_user_handler;
   st->user_callback.data = NULL;
   for (i=0;i<16;i++)
      st->speex_callbacks[i].func = NULL;

   st->voc_m1=st->voc_m2=st->voc_mean=0;
   st->voc_offset=0;
   st->dtx_enabled=0;
#ifdef ENABLE_VALGRIND
   VALGRIND_MAKE_READABLE(st, (st->stack-(char*)st));
#endif
   return st;
}

void nb_decoder_destroy(void *state)
{
   DecState *st;
   st=(DecState*)state;
   
#if !(defined(VAR_ARRAYS) || defined (USE_ALLOCA))
   speex_free_scratch(st->stack);
#endif

   speex_free (st->inBuf);
   speex_free (st->excBuf);
   speex_free (st->innov);
   speex_free (st->interp_qlpc);
   speex_free (st->qlsp);
   speex_free (st->old_qlsp);
   speex_free (st->interp_qlsp);
   speex_free (st->mem_sp);
   speex_free (st->comb_mem);
   speex_free (st->pi_gain);

   speex_free(state);
}

#define median3(a, b, c)	((a) < (b) ? ((b) < (c) ? (b) : ((a) < (c) ? (c) : (a))) : ((c) < (b) ? (b) : ((c) < (a) ? (c) : (a))))

#ifdef FIXED_POINT
const spx_word16_t attenuation[10] = {32767, 31483, 27923, 22861, 17278, 12055, 7764, 4616, 2533, 1283};
#else
const spx_word16_t attenuation[10] = {1., 0.961, 0.852, 0.698, 0.527, 0.368, 0.237, 0.141, 0.077, 0.039};

#endif

static void nb_decode_lost(DecState *st, spx_word16_t *out, char *stack)
{
   int i, sub;
   VARDECL(spx_coef_t *awk1);
   VARDECL(spx_coef_t *awk2);
   VARDECL(spx_coef_t *awk3);
   spx_word16_t pitch_gain;
   spx_word16_t fact;
   spx_word16_t gain_med;
   spx_word16_t innov_gain;
   
   if (st->count_lost<10)
      fact = attenuation[st->count_lost];
   else
      fact = 0;

   gain_med = median3(st->pitch_gain_buf[0], st->pitch_gain_buf[1], st->pitch_gain_buf[2]);
   if (gain_med < st->last_pitch_gain)
      st->last_pitch_gain = gain_med;
   
#ifdef FIXED_POINT
   pitch_gain = st->last_pitch_gain;
   if (pitch_gain>62)
      pitch_gain = 62;
   pitch_gain = SHL(pitch_gain, 9);
#else   
   pitch_gain = GAIN_SCALING_1*st->last_pitch_gain;
   if (pitch_gain>.95)
      pitch_gain=.95;
#endif

   pitch_gain = MULT16_16_Q15(fact,pitch_gain) + VERY_SMALL;

   /* Shift all buffers by one frame */
   /*speex_move(st->inBuf, st->inBuf+st->frameSize, (st->bufSize-st->frameSize)*sizeof(spx_sig_t));*/
   speex_move(st->excBuf, st->excBuf+st->frameSize, (st->max_pitch + 1)*sizeof(spx_sig_t));

   ALLOC(awk1, (st->lpcSize+1), spx_coef_t);
   ALLOC(awk2, (st->lpcSize+1), spx_coef_t);
   ALLOC(awk3, (st->lpcSize+1), spx_coef_t);

   for (sub=0;sub<st->nbSubframes;sub++)
   {
      int offset;
      spx_sig_t *sp, *exc;
      /* Offset relative to start of frame */
      offset = st->subframeSize*sub;
      /* Original signal */
      sp=st->frame+offset;
      /* Excitation */
      exc=st->exc+offset;
      /* Excitation after post-filter*/

      /* Calculate perceptually enhanced LPC filter */
      if (st->lpc_enh_enabled)
      {
         spx_word16_t k1,k2,k3;
         if (st->submodes[st->submodeID] != NULL)
         {
            k1=SUBMODE(lpc_enh_k1);
            k2=SUBMODE(lpc_enh_k2);
            k3=SUBMODE(lpc_enh_k3);
         } else {
            k1=k2=.7*GAMMA_SCALING;
            k3=.0;
         }
         bw_lpc(k1, st->interp_qlpc, awk1, st->lpcSize);
         bw_lpc(k2, st->interp_qlpc, awk2, st->lpcSize);
         bw_lpc(k3, st->interp_qlpc, awk3, st->lpcSize);
      }
        
      /* Make up a plausible excitation */
      /* FIXME: THIS CAN BE IMPROVED */
      /*if (pitch_gain>.95)
        pitch_gain=.95;*/
      innov_gain = compute_rms(st->innov, st->frameSize);
      for (i=0;i<st->subframeSize;i++)
      {
         exc[i]= MULT16_32_Q15(pitch_gain, (exc[i-st->last_pitch]+VERY_SMALL)) + 
               MULT16_32_Q15(fact, MULT16_32_Q15(sqrt(SHL(Q15ONE,15)-SHL(pitch_gain,15)),speex_rand(innov_gain, &st->seed)));
      }
      
      for (i=0;i<st->subframeSize;i++)
         sp[i]=exc[i];
      
      /* Signal synthesis */
      if (st->lpc_enh_enabled)
      {
         filter_mem2(sp, awk2, awk1, sp, st->subframeSize, st->lpcSize, 
                     st->mem_sp+st->lpcSize);
         filter_mem2(sp, awk3, st->interp_qlpc, sp, st->subframeSize, st->lpcSize, 
                     st->mem_sp);
      } else {
         for (i=0;i<st->lpcSize;i++)
            st->mem_sp[st->lpcSize+i] = 0;
         iir_mem2(sp, st->interp_qlpc, sp, st->subframeSize, st->lpcSize, 
                     st->mem_sp);
      }      
   }

   for (i=0;i<st->frameSize;i++)
   {
      spx_word32_t sig = PSHR32(st->frame[i],SIG_SHIFT);
      if (sig>32767)
         sig = 32767;
      if (sig<-32767)
         sig = -32767;
     out[i]=sig;
   }
   
   st->first = 0;
   st->count_lost++;
   st->pitch_gain_buf[st->pitch_gain_buf_idx++] = PSHR(pitch_gain,9);
   if (st->pitch_gain_buf_idx > 2) /* rollover */
      st->pitch_gain_buf_idx = 0;
}

int nb_decode(void *state, SpeexBits *bits, void *vout)
{
   DecState *st;
   int i, sub;
   int pitch;
   spx_word16_t pitch_gain[3];
   spx_word32_t ol_gain=0;
   int ol_pitch=0;
   spx_word16_t ol_pitch_coef=0;