123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275 |
- #ifdef HAVE_CONFIG_H
- #include "config.h"
- #endif
- #include "vbr.h"
- #include <math.h>
- #define sqr(x) ((x)*(x))
- #define MIN_ENERGY 6000
- #define NOISE_POW .3
- #ifndef DISABLE_VBR
- const float vbr_nb_thresh[9][11]={
- {-1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f},
- { 4.0f, 2.5f, 2.0f, 1.2f, 0.5f, 0.0f, -0.5f, -0.7f, -0.8f, -0.9f, -1.0f},
- {10.0f, 6.5f, 5.2f, 4.5f, 3.9f, 3.5f, 3.0f, 2.5f, 2.3f, 1.8f, 1.0f},
- {11.0f, 8.8f, 7.5f, 6.5f, 5.0f, 3.9f, 3.9f, 3.9f, 3.5f, 3.0f, 1.0f},
- {11.0f, 11.0f, 9.9f, 8.5f, 7.0f, 6.0f, 4.5f, 4.0f, 4.0f, 4.0f, 2.0f},
- {11.0f, 11.0f, 11.0f, 11.0f, 9.5f, 8.5f, 8.0f, 7.0f, 6.0f, 5.0f, 3.0f},
- {11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 9.5f, 8.5f, 7.0f, 6.0f, 5.0f},
- {11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 9.8f, 9.5f, 7.5f},
- { 7.0f, 4.5f, 3.7f, 3.0f, 2.5f, 2.0f, 1.8f, 1.5f, 1.0f, 0.0f, 0.0f}
- };
- const float vbr_hb_thresh[5][11]={
- {-1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f},
- {-1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f},
- {11.0f, 11.0f, 9.5f, 8.5f, 7.5f, 6.0f, 5.0f, 3.9f, 3.0f, 2.0f, 1.0f},
- {11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 9.5f, 8.7f, 7.8f, 7.0f, 6.5f, 4.0f},
- {11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 11.0f, 9.8f, 7.5f, 5.5f}
- };
- const float vbr_uhb_thresh[2][11]={
- {-1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f, -1.0f},
- { 3.9f, 2.5f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, -1.0f}
- };
- void vbr_init(VBRState *vbr)
- {
- int i;
- vbr->average_energy=0;
- vbr->last_energy=1;
- vbr->accum_sum=0;
- vbr->energy_alpha=.1;
- vbr->soft_pitch=0;
- vbr->last_pitch_coef=0;
- vbr->last_quality=0;
- vbr->noise_accum = .05*pow(MIN_ENERGY, NOISE_POW);
- vbr->noise_accum_count=.05;
- vbr->noise_level=vbr->noise_accum/vbr->noise_accum_count;
- vbr->consec_noise=0;
- for (i=0;i<VBR_MEMORY_SIZE;i++)
- vbr->last_log_energy[i] = log(MIN_ENERGY);
- }
- float vbr_analysis(VBRState *vbr, spx_word16_t *sig, int len, int pitch, float pitch_coef)
- {
- int i;
- float ener=0, ener1=0, ener2=0;
- float qual=7;
- int va;
- float log_energy;
- float non_st=0;
- float voicing;
- float pow_ener;
- for (i=0;i<len>>1;i++)
- ener1 += ((float)sig[i])*sig[i];
- for (i=len>>1;i<len;i++)
- ener2 += ((float)sig[i])*sig[i];
- ener=ener1+ener2;
- log_energy = log(ener+MIN_ENERGY);
- for (i=0;i<VBR_MEMORY_SIZE;i++)
- non_st += sqr(log_energy-vbr->last_log_energy[i]);
- non_st = non_st/(30*VBR_MEMORY_SIZE);
- if (non_st>1)
- non_st=1;
- voicing = 3*(pitch_coef-.4)*fabs(pitch_coef-.4);
- vbr->average_energy = (1-vbr->energy_alpha)*vbr->average_energy + vbr->energy_alpha*ener;
- vbr->noise_level=vbr->noise_accum/vbr->noise_accum_count;
- pow_ener = pow(ener,NOISE_POW);
- if (vbr->noise_accum_count<.06 && ener>MIN_ENERGY)
- vbr->noise_accum = .05*pow_ener;
- if ((voicing<.3 && non_st < .2 && pow_ener < 1.2*vbr->noise_level)
- || (voicing<.3 && non_st < .05 && pow_ener < 1.5*vbr->noise_level)
- || (voicing<.4 && non_st < .05 && pow_ener < 1.2*vbr->noise_level)
- || (voicing<0 && non_st < .05))
- {
- float tmp;
- va = 0;
- vbr->consec_noise++;
- if (pow_ener > 3*vbr->noise_level)
- tmp = 3*vbr->noise_level;
- else
- tmp = pow_ener;
- if (vbr->consec_noise>=4)
- {
- vbr->noise_accum = .95*vbr->noise_accum + .05*tmp;
- vbr->noise_accum_count = .95*vbr->noise_accum_count + .05;
- }
- } else {
- va = 1;
- vbr->consec_noise=0;
- }
- if (pow_ener < vbr->noise_level && ener>MIN_ENERGY)
- {
- vbr->noise_accum = .95*vbr->noise_accum + .05*pow_ener;
- vbr->noise_accum_count = .95*vbr->noise_accum_count + .05;
- }
-
- if (ener < 30000)
- {
- qual -= .7;
- if (ener < 10000)
- qual-=.7;
- if (ener < 3000)
- qual-=.7;
- } else {
- float short_diff, long_diff;
- short_diff = log((ener+1)/(1+vbr->last_energy));
- long_diff = log((ener+1)/(1+vbr->average_energy));
-
- if (long_diff<-5)
- long_diff=-5;
- if (long_diff>2)
- long_diff=2;
- if (long_diff>0)
- qual += .6*long_diff;
- if (long_diff<0)
- qual += .5*long_diff;
- if (short_diff>0)
- {
- if (short_diff>5)
- short_diff=5;
- qual += .5*short_diff;
- }
-
- if (ener2 > 1.6*ener1)
- qual += .5;
- }
- vbr->last_energy = ener;
- vbr->soft_pitch = .6*vbr->soft_pitch + .4*pitch_coef;
- qual += 2.2*((pitch_coef-.4) + (vbr->soft_pitch-.4));
- if (qual < vbr->last_quality)
- qual = .5*qual + .5*vbr->last_quality;
- if (qual<4)
- qual=4;
- if (qual>10)
- qual=10;
-
-
- if (vbr->consec_noise>=3)
- qual=4;
- if (vbr->consec_noise)
- qual -= 1.0 * (log(3.0 + vbr->consec_noise)-log(3));
- if (qual<0)
- qual=0;
-
- if (ener<60000)
- {
- if (vbr->consec_noise>2)
- qual-=0.5*(log(3.0 + vbr->consec_noise)-log(3));
- if (ener<10000&&vbr->consec_noise>2)
- qual-=0.5*(log(3.0 + vbr->consec_noise)-log(3));
- if (qual<0)
- qual=0;
- qual += .3*log(.0001+ener/60000.0);
- }
- if (qual<-1)
- qual=-1;
-
- vbr->last_pitch_coef = pitch_coef;
- vbr->last_quality = qual;
- for (i=VBR_MEMORY_SIZE-1;i>0;i--)
- vbr->last_log_energy[i] = vbr->last_log_energy[i-1];
- vbr->last_log_energy[0] = log_energy;
-
- return qual;
- }
- void vbr_destroy(VBRState *vbr)
- {
- }
- #endif
|