gprune_none.c

julius version 4.12.about sound recognition.

/**
 * @file   gprune_none.c
 * 
 * <JA>
 * @brief  寒圭ガウス尸邵纷换: Gaussian pruning 痰し
 *
 * gprune_none()は寒圭ガウス尸邵礁圭の纷换ル〖チンの办つですˉ
 * pruningなどを乖なわずに链てのGauss尸邵について叫蜗澄唯を滇めますˉ
 * tied-mixtureでないモデルではこの簇眶がデフォルトで脱いられますˉ
 * また "-gprune none" を回年することでも联买することができますˉ
 *
 * outprob_init() によって gprune_none() が簇眶ポインタ @a compute_gaussset に
 * セットされ·それが calc_tied_mix() またはcalc_mix() から钙び叫されますˉ
 * </JA>
 * 
 * <EN>
 * @brief  Calculate probability of a set of Gaussian densities (no pruning)
 *
 * gprune_none() is one of the functions to compute output probability of
 * a set of Gaussian densities.  This function does no pruning, just computing
 * each Gaussian density one by one.  For non tied-mixture model, this function
 * is used as default.  Specifying "-gprune none" at runtime can also select
 * this function.
 *
 * gprune_none() will be used by calling outprob_init() to set its pointer
 * to the global variable @a compute_gaussset.  Then it will be called from
 * calc_tied_mix() or calc_mix().
 * </EN>
 * 
 * @author Akinobu LEE
 * @date   Thu Feb 17 05:09:46 2005
 *
 * $Revision: 1.4 $
 * 
 */
/*
 * Copyright (c) 1991-2007 Kawahara Lab., Kyoto University
 * Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and Technology
 * Copyright (c) 2005-2007 Julius project team, Nagoya Institute of Technology
 * All rights reserved
 */

#include <sent/stddefs.h>
#include <sent/htk_hmm.h>
#include <sent/htk_param.h>
#include <sent/hmm.h>
#include <sent/hmm_calc.h>

/** 
 * Calculate probability of a Gaussian density against input
 * vector on OP_vec.
 * 
 * @param wrk [i/o] HMM computation work area
 * @param binfo [in] a Gaussian density
 * 
 * @return the output log probability.
 */
LOGPROB
compute_g_base(HMMWork *wrk, HTK_HMM_Dens *binfo)
{
  VECT tmp, x;
  VECT *mean;
  VECT *var;
  VECT *vec = wrk->OP_vec;
  short veclen = wrk->OP_veclen;

  if (binfo == NULL) return(LOG_ZERO);
  mean = binfo->mean;
  var = binfo->var->vec;
  tmp = binfo->gconst;
  for (; veclen > 0; veclen--) {
#ifdef ENABLE_MSD
    if (*vec == LZERO) {
      vec++;
      continue;
    }
#endif
    x = *(vec++) - *(mean++);
    tmp += x * x * *(var++);
  }
  return(tmp * -0.5);
}

/** 
 * Initialize and setup work area for Gaussian computation
 * 
 * @param wrk [i/o] HMM computation work area
 * 
 * @return TRUE on success, FALSE on failure.
 */
boolean
gprune_none_init(HMMWork *wrk)
{
  /* maximum Gaussian set size = maximum mixture size * nstream */
  wrk->OP_calced_maxnum = wrk->OP_hmminfo->maxmixturenum * wrk->OP_nstream;
  wrk->OP_calced_score = (LOGPROB *)mymalloc(sizeof(LOGPROB) * wrk->OP_calced_maxnum);
  wrk->OP_calced_id = (int *)mymalloc(sizeof(int) * wrk->OP_calced_maxnum);
  /* force gprune_num to the max number */
  wrk->OP_gprune_num = wrk->OP_calced_maxnum;
  return TRUE;
}

/**
 * Free gprune_none related work area.
 * 
 * @param wrk [i/o] HMM computation work area
 * 
 */
void
gprune_none_free(HMMWork *wrk)
{
  free(wrk->OP_calced_score);
  free(wrk->OP_calced_id);
}

/** 
 * @brief  Compute a set of Gaussians with no pruning
 *
 * The calculated scores will be stored to OP_calced_score, with its
 * corresponding mixture id to OP_calced_id. 
 * The number of calculated mixtures is also stored in OP_calced_num.
 * 
 * This can be called from calc_tied_mix() or calc_mix().
 * 
 * @param wrk [i/o] HMM computation work area
 * @param g [in] set of Gaussian densities to compute the output probability.
 * @param num [in] length of above
 * @param last_id [in] ID list of N-best mixture in previous input frame,
 * or NULL if not exist
 * @param lnum [in] length of last_id
 */
void
gprune_none(HMMWork *wrk, HTK_HMM_Dens **g, int num, int *last_id, int lnum)
{
  int i;
  HTK_HMM_Dens *dens;
  LOGPROB *prob = wrk->OP_calced_score;
  int *id = wrk->OP_calced_id;
#ifdef ENABLE_MSD
  int valid_dim;
  int calced_num;
#endif

#ifdef ENABLE_MSD

  valid_dim = 0;
  for(i=0; i<wrk->OP_veclen; i++) {
    if (wrk->OP_vec[i] != LZERO) valid_dim++;
  }
  calced_num = 0;
  for(i=0; i<num; i++) {
    dens = *(g++);
    if (dens->meanlen != valid_dim) continue;
    if (dens->meanlen == 0) {
      *(prob++) = 0.0;
    } else {
      *(prob++) = compute_g_base(wrk, dens);
    }
    *(id++) = i;
    calced_num++;
  }
  if (calced_num == 0) {
    jlog("Error: MSD: input data dim = %d / %d, but no Gaussian defined for it\n", valid_dim, wrk->OP_veclen);
    jlog("Error: MSD: Gaussian dimensions in this mixture:");
    for(i=0;i<num;i++) {
      jlog(" %d", g[i]->meanlen);
    }
    jlog("\n");
  }
  wrk->OP_calced_num = calced_num;

#else

  for(i=0; i<num; i++) {
    dens = *(g++);
    *(prob++) = compute_g_base(wrk, dens);
    *(id++) = i;
  }
  wrk->OP_calced_num = num;

#endif
}