ht_param_0.cc

这是处理语音信号的程序

// file: ht_param_0.cc
//
// function to read data from parameter file
//

// system include files
//
#include <string.h>

// isip include files
//
#include "hmm_train.h"
#include "hmm_train_constants.h"

// method: read_params_cc
//
// arguments:
//  FILE* fp : (input) file pointer to params file
//  char_1* mono_file : (output) name of file containing monophones
//  char_1* trans_file : (input) name of transitions data file
//  char_1* state_file : (output) name of file containing updated states data
//  char_1* new_trans_file : (output) name of updated transitions data file
//  char_1* new_state_file : (input) name of file containing states data
//  char_1* batch_stats_file : (input) name of file containing viterbi
//                                     accumulators
//  char_1* acc_list_file : (input) name of file containing list of viterbi
//                                     accumulators
//  char_1* model_file : (output) name of file containing models
//  char_1* phone_file : (output) name of file containing triphones
//  char_1* lex_file : (output) name of lexicon file
//  int_4& align_mode : (output) alignment mode (state/model)
//  int_4& input_mode : (output) input mode (ascii/binary)
//  int_4 level : (input) number of levels in the search hierarchy
//  char_1* input_file : (output) name of the input file
//  char_1* mlf_file : (output) name of the mlf file
//  char_1* output_file : (output) name of the output file
//  int_4& phn : (output) size of phonetic context
//  int_4& nbst : (output) n-best word hyps
//  float_8*& beam : (output) beam widths for pruning
//  int_4& mapmi : (output) maximum number of active traces
//  logical_1& mlf_mode_flag : (output) if this is model mlf or word mlf
//  int_4& context_mode : (output) context_mode
//  float_8*& var_floor: (output) variance floor
//  logical_1& op_out: (output) the flag of output option
//  logical_1& output_mode: (output) format for the output states file
//  logical_1& train_mode: (output) training mode (batch/normal)
//  char_1* sph_file: (output) name of the specail phone list file
//  char_1* state_occ_file: (output) name of the state occupancy file
//  char_1* occ_mode: (output) whether or not we print occupancies
//
// return: a logical flag to indicate success
//
logical_1 read_params_cc(FILE* fp_a, char_1* mono_file_a, char_1* trans_file_a,
			 char_1* state_file_a, char_1* new_trans_file_a,
			 char_1* new_state_file_a, char_1* batch_stats_file_a,
			 char_1* acc_list_file_a, char_1* model_file_a,
			 char_1* phone_file_a, char_1* lex_file_a,
			 int_4& align_mode_a, int_4& input_mode_a,
			 int_4 level_a, char_1* input_file_a,
			 char_1* mlf_file_a, char_1* output_file_a,
			 int_4& phn_a, int_4& nbst_a, float_8*& beam_a,
			 int_4& mapmi_a, logical_1& mlf_mode_a,
			 int_4& context_mode_a, float_8& var_floor_a,
			 logical_1& op_out_a, logical_1& output_mode_a,
			 logical_1& train_mode_a, char_1* sph_file_a,
			 char_1* state_occ_file_a, logical_1& occ_mode_a) {
  
  // string to hold temporary data
  //
  char_1* tmp = new char_1[ISIP_MAX_STRING_LENGTH];
  tmp = (char_1*)memset(tmp, (int_4)0, ISIP_MAX_STRING_LENGTH);
  char_1* tag_str = new char_1[ISIP_MAX_STRING_LENGTH];
  tag_str = (char_1*)memset(tag_str, (int_4)0, ISIP_MAX_STRING_LENGTH);
  char_1* tmp1 = new char_1[ISIP_MAX_STRING_LENGTH];
  tmp1 = (char_1*)memset(tmp1, (int_4)0, ISIP_MAX_STRING_LENGTH);
  char_1* tmp2 = new char_1[ISIP_MAX_STRING_LENGTH];
  tmp2 = (char_1*)memset(tmp2, (int_4)0, ISIP_MAX_STRING_LENGTH);
  char_1* tmp3 = new char_1[ISIP_MAX_STRING_LENGTH];
  tmp3 = (char_1*)memset(tmp3, (int_4)0, ISIP_MAX_STRING_LENGTH);
  char_1* tmp4 = new char_1[ISIP_MAX_STRING_LENGTH];
  tmp4 = (char_1*)memset(tmp4, (int_4)0, ISIP_MAX_STRING_LENGTH);
  char_1* tmp5 = new char_1[ISIP_MAX_STRING_LENGTH];
  tmp5 = (char_1*)memset(tmp5, (int_4)0, ISIP_MAX_STRING_LENGTH);

  // read data from file
  //
  while (fgets((char*)tmp, ISIP_MAX_STRING_LENGTH - 1, fp_a) != (char*)NULL) {
    
    // ignore comment lines 
    //
    if (tmp[0] == (char_1)'#') {
      
      // do nothing
      //
      continue;
    }

    // get the first string which should be the tag string. if the line is
    // empty then 0 should be returned
    //
    if (sscanf((char*)tmp, "%s", tag_str) == 0) {
      continue;
    }
    
    // read name of monophones file
    //
    if (strcmp((char*)tag_str, (char*)HT_MONOPHONE_FILE) == 0) {
  
      
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)mono_file_a, (char*)tmp3); 
	expand_filename_cc(mono_file_a);
      }
    }

    //read the input file
    //
   else if (strcmp((char*)tag_str, HT_INPUT_FILE ) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)input_file_a, (char*)tmp3); 
	expand_filename_cc(input_file_a);
      }
    }

    // read the mlf file
    //
    else if (strcmp((char*)tag_str,  HT_MLF_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)mlf_file_a, (char*)tmp3); 
	expand_filename_cc(mlf_file_a);
      }
    }
    
    // read the output file 
    //
    else if (strcmp((char*)tag_str,HT_OUTPUT_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)output_file_a, (char*)tmp3); 
	expand_filename_cc(output_file_a);
	op_out_a = ISIP_TRUE;
      }
    }

    //set the phonetic context size
    //
    else if (strcmp(HT_CONTEXT_SIZE, (char*)tag_str) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	// curently assumes 3 anyways
	//
	//phn_a = (int_4)atoi(tmp3);
      }
    }
  
    // read the value for n-best
    //
    else if (strcmp(HT_NBEST, (char*)tag_str) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	nbst_a = (int_4)atoi((char*)tmp3);
      }
    }

    // set the beam widths if beam pruning is allowed
    //
    else if (strcmp(HT_BEAM_WIDTH, (char*)tag_str) == 0) {
      if (sscanf((char*)tmp, "%s%s%s%s%s", tmp1, tmp2, tmp3, tmp4, tmp5) == 5) {
	  beam_a[0] = -(float_8)fabs((float_8)atof((char*)tmp3));
	  beam_a[1] = -(float_8)fabs((float_8)atof((char*)tmp4));
	  beam_a[2] = -(float_8)fabs((float_8)atof((char*)tmp5));
      }
    }

    // set the variance floor for training
    //
    else if (strcmp((char*)HT_VARIANCE_FLOOR_NAME, (char*)tag_str) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	var_floor_a = (float_8)atof((char*)tmp3);
      }
    }

    // set the limit on maximum active phone model instances if
    // specified
    //
    else if (strcmp(HT_MAPMI, (char*)tag_str) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	mapmi_a = (int_4)atof((char*)tmp3);
	if (mapmi_a < (int_4)0) {
	  mapmi_a = (int_4)0;
	}
      }
    }
    
    // read name of transitions data file
    //
    else if (strcmp((char*)tag_str, HT_TRANSITIONS_FILE ) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)trans_file_a, (char*)tmp3); 
	expand_filename_cc(trans_file_a);
      }
    }
    
    // read name of states data file
    //
    else if (strcmp((char*)tag_str,  HT_STATES_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)state_file_a, (char*)tmp3); 
	expand_filename_cc(state_file_a);
      }
    }

    // read name of updated transitions data file
    //
    else if (strcmp((char*)tag_str, HT_UPDATED_TRANS_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)new_trans_file_a, (char*)tmp3); 
	expand_filename_cc(new_trans_file_a);
      }
    }
    
    // read name of updated states data file
    //
    else if (strcmp((char*)tag_str, HT_UPDATED_STATE_FILE ) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)new_state_file_a, (char*)tmp3); 
	expand_filename_cc(new_state_file_a);
      }
    }
    
    // read name of batch stats file
    //
    else if (strcmp((char*)tag_str, HT_BATCH_STATS_FILE ) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)batch_stats_file_a, (char*)tmp3); 
	expand_filename_cc(batch_stats_file_a);
      }
    }
    
    // read name of file containing the list of viterbi accumulators
    //
    else if (strcmp((char*)tag_str, HT_ACC_LIST_FILE ) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)acc_list_file_a, (char*)tmp3); 
	expand_filename_cc(acc_list_file_a);
      }
    }
    
    // read name of HMM models data file
    //
    else if (strcmp((char*)tag_str, HT_MODELS_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)model_file_a, (char*)tmp3); 
	expand_filename_cc(model_file_a);
      }
    }
    
    // read name of triphones data file
    //
    else if (strcmp((char*)tag_str,  HT_PHONES_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)phone_file_a, (char*)tmp3); 
	expand_filename_cc(phone_file_a);
      }
    }
    
    // read name of lexicon data file
    //
    else if (strcmp((char*)tag_str, HT_LEXICON_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)lex_file_a, (char*)tmp3); 
	expand_filename_cc(lex_file_a);
      }
    }

 
    // read context mode
    //
    else if (strcmp((char*)tag_str, HT_CONTEXT_MODE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {

	if(strcmp((char*) HT_WORD_INTERNAL_MODE, (char*)tmp3) ==0){
	  context_mode_a = 2;
	}
	else if(strcmp((char*)HT_MONO_PHONE_MODE, (char*)tmp3) ==0){
	  context_mode_a = 1;
	}
	else if(strcmp((char*)HT_CROSS_WORD_MODE, (char*)tmp3) ==0){
	  context_mode_a = 3;
	}
	else {
	  fprintf(stdout, "\nht_param_0.cc:  context mode error\n\n");
	  
	  // exit gracefully 
	  //
	  exit(ISIP_PROTO_ERROR);
	}
      }    
    }

    // read alignment mode
    //
    else if (strcmp((char*)tag_str, (char*)HT_ALIGN_MODE_NAME) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {

	if (strcmp((char*)tmp3, (char*)HT_MODEL_ALIGN_NAME) == 0) {
	  align_mode_a = HT_MODEL_ALIGN;
	}
	else if (strcmp((char*)tmp3, (char*)HT_STATE_ALIGN_NAME) == 0) {
	  align_mode_a = HT_STATE_ALIGN;
	}
	else {
	  fprintf(stdout, "alignment mode \"%s\" not recognized\n", tmp3);
	
	  // exit disgracefully
	  //
	  exit(ISIP_PROTO_ERROR);
	}
      }
    }

    // read input mode
    //
    else if (strcmp((char*)tag_str, (char*)HT_INPUT_MODE_NAME) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	if (strcmp((char*)tmp3, (char*)HT_ASCII) == 0) {
	  input_mode_a = HT_ASCII_MODE;
	}
	else if (strcmp((char*)tmp3, (char*)HT_BINARY) == 0) {
	  input_mode_a = HT_BINARY_MODE;
	}
      }
    }
    
    // read input mode
    //
    else if (strcmp((char*)tag_str, (char*)HT_OUTPUT_MODE_NAME) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	if (strcmp((char*)tmp3, (char*)HT_ASCII) == 0) {
	  output_mode_a = HT_ASCII_MODE;
	}
	else if (strcmp((char*)tmp3, (char*)HT_BINARY) == 0) {
	  output_mode_a = HT_BINARY_MODE;
	}
      }
    }

    // read mlf mode
    //
    else if (strcmp((char*)tag_str, (char*)HT_MLF_MODE_NAME) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	if (strcmp((char*)tmp3, (char*)HT_MODEL_MLF) == 0) {
	  mlf_mode_a = HT_MODEL_TRANS;
	}
	else if (strcmp((char*)tmp3, (char*)HT_WORD_MLF) == 0) {
	  mlf_mode_a = HT_WORD_TRANS;
	}
      }
    }
    
    // read training mode
    //
    else if (strcmp((char*)tag_str, (char*)HT_TRAIN_MODE_NAME) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	if (strcmp((char*)tmp3, (char*)HT_NORMAL_MODE) == 0) {
	  train_mode_a = HT_NORMAL;
	}
	else if (strcmp((char*)tmp3, (char*)HT_BATCH_MODE) == 0) {
	  train_mode_a = HT_BATCH;
	}
	else if (strcmp((char*)tmp3, (char*)HT_COMBINE_MODE) == 0) {
	  train_mode_a = HT_COMBINE;
	}
      }
    }

    // read the name of the special phone list file
    //    
    else if (strcmp((char*)tag_str,(char*)HT_SPH_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)sph_file_a, (char*)tmp3); 
	expand_filename_cc(sph_file_a);
      }
    }
    
    // read occupancy mode and/or filename
    //
    else if (strcmp((char*)tag_str, HT_STATE_OCCUPANCY) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	if (strcmp((char*)tmp3, (char*)HT_OPT_ON_NAME) == 0) {
	  occ_mode_a = HT_OPT_ON;
	}
	else if (strcmp((char*)tmp3, (char*)HT_OPT_OFF_NAME) == 0) {
	  occ_mode_a = HT_OPT_OFF;
	}
      }
    }

    else if (strcmp((char*)tag_str, (char*)HT_STATE_OCC_FILE) == 0) {
      if (sscanf((char*)tmp, "%s%s%s", tmp1, tmp2, tmp3) == 3) {
	strcpy((char*)state_occ_file_a, (char*)tmp3); 
	expand_filename_cc(state_occ_file_a);
      }
    }

    // otherwise ignore the line
    //
    else {
    }
  }
  
  // delete memory
  //
  delete [] tmp;
  delete [] tag_str;
  delete [] tmp1;
  delete [] tmp2;
  delete [] tmp3;
  delete [] tmp4;
  delete [] tmp5;
  
  // return gracefully
  //
  return(ISIP_TRUE);
}