tr_lat_read_0.cc

这是处理语音信号的程序

// file: tr_lat_read_0.cc
//

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

// isip include files
//
#include "train_lattice.h"
#include "train_lattice_constants.h"

// method: read_lattice_cc
//
// arguments:
//  FILE* fp_in_a: (input) input Train_Lattice file
//  int_4* Train_Word_table : (input) the hash table of Train_Word indices
//
// return: a logical_1 indicating status
//
// this method reads in the Train_Lattice from a file
//
logical_1 Train_Lattice::read_lattice_cc(FILE* fp_in_a, Train_Hash_table* words_a) {

  // define some local variables
  //
  char_1* buffer = new char_1[ISIP_MAX_STRING_LENGTH];
  char_1* buffer_pos = buffer;
  char_1* temp_buf = (char_1*)NULL;
  char_1* delimiter = (char_1*)"=";

  float_4 time = (float_4)0;
  int_4 num_nodes = (int_4)0;
  int_4 num_arcs = (int_4)0;

  int_4 frame_index = (int_4)0;
  int_4 node_index = (int_4)0;
  int_4 arc_count = (int_4)0;

  int_4 start_node = (int_4)0;
  int_4 end_node = (int_4)0;

  logical_1 read_words = ISIP_TRUE;
  int_4 pron_var = (int_4)0;

  float_8 ac_like = (float_8)0;
  float_8 lm_score = (float_8)0;
  float_4 lm_scale = (float_4)0;
  float_4 word_penalty = (float_4)0;

  Train_Word* word = (Train_Word*)NULL;
  Train_Hash_cell* hcell = (Train_Hash_cell*)NULL;
  Train_Lattice_node** lnodes = (Train_Lattice_node**)NULL;;
  
  // define memory
  //
  memset((char*)buffer, 0, ISIP_MAX_STRING_LENGTH);
  
  // read in the lines till we get to reading the link/arc information
  // read the Train_Lattice generation specific information like utterance
  // number, word insertion penalty, lexicon, and Train_Model set
  // also read the node information
  //
  while (fgets((char*)buffer_pos, ISIP_MAX_STRING_LENGTH, fp_in_a) &&
	 (*buffer_pos != 'J')) {
    
    // find the delimited value
    //
    temp_buf = (char_1*)strtok((char*)buffer_pos, (char*)delimiter);
    
    // parse this string
    //
    while (temp_buf != (char_1*)NULL) {
      
      // move the pointer to first non-whitespace
      //
      while (isspace((char)(*temp_buf))) {
	temp_buf++;
      }
      
      // read the Train_Lattice generation information
      // version number of generation program (typically HVite)
      //
      if (strcmp((char*)temp_buf, "VERSION") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // utterance represented by the Train_Lattice
      //
      else if (strcmp((char*)temp_buf, "UTTERANCE") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	set_utterance_cc((char_1*)temp_buf);
      }
      
      // word insertion penalty used during Train_Lattice generation
      //
      else if (strcmp((char*)temp_buf, "wdpenalty") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	word_penalty = atof((char*)temp_buf);
	set_word_penalty_cc(word_penalty);
      }
      
      // language Train_Model scaling used during Train_Lattice generation
      //
      else if (strcmp((char*)temp_buf, "lmscale") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	lm_scale = atof((char*)temp_buf);
	set_lm_scale_cc(lm_scale);
      }
      
      // language Train_Model used during Train_Lattice generation
      //
      else if (strcmp((char*)temp_buf, "vocab") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	set_lang_model_cc((char_1*)temp_buf);
      }
      
      // Train_Model set used during Train_Lattice generation
      //
      else if (strcmp((char*)temp_buf, "hmms") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	set_model_set_cc((char_1*)temp_buf);
      }
      
      // number of nodes in the Train_Lattice
      //
      else if (strcmp((char*)temp_buf, "N") == 0) {
	
	// do not confuse with alphabet N which could be a word
	//
	if (read_words == ISIP_TRUE) {

	  // set the number of Train_Lattice nodes
	  //
	  temp_buf = (char_1*)strtok((char*)NULL, " ");
	  num_nodes = atoi((char*)temp_buf);
	  set_num_nodes_cc(num_nodes);
	  
	  // initialize the Train_Lattice nodes table
	  //
	  lnodes = new Train_Lattice_node*[num_nodes];
	  init_nodes_cc(lnodes);
  	}

	// otherwise ignore this word and move on
	//
	else {
	  temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	}
      }
      
      // number of arcs in the Train_Lattice
      //
      else if (strcmp((char*)temp_buf, "L") == 0) {
	
	// do not confuse with alphabet L which could be a word
	//
	if (read_words == ISIP_TRUE) {

	  // set the number of arcs
	  //
	  temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	  num_arcs = atoi((char*)temp_buf);
	  set_num_arcs_cc(num_arcs);
	}

	// otherwise ignore this word and move on
	//
	else {
	  temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	}
      }
      
      // read the node information (node index, end time, word etc.)
      //
      else if (strcmp((char*)temp_buf, "I") == 0) {
	
	// do not confuse with alphabet I which could be a word
	//
	if (read_words == ISIP_TRUE) {

	  // set the node index
	  //
	  temp_buf = (char_1*)strtok((char*)NULL, " ");
	  node_index = atoi((char*)temp_buf);
	  lnodes[node_index]->set_node_index_cc(node_index);
	}

	// otherwise ignore this word and move on
	//
	else {
	  temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	}
      }

      // read the frame information for this node
      //
      else if (strcmp((char*)temp_buf, "t") == 0) {
	
	// do not confuse with alphabet t which could be a word
	//
	if (read_words == ISIP_TRUE) {

	  // set the frame index
	  //
	  temp_buf = (char_1*)strtok((char*)NULL, " ");
	  time = atof((char*)temp_buf);
	  frame_index = (int_4)rint(time * TRAIN_LATTICE_FRAMES_PER_SEC);
	  lnodes[node_index]->set_frame_index_cc(frame_index);
	}

	// otherwise ignore this word and move on
	//
	else {
	  temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	}
      }
      
      // read the word index of the word represented by the node
      //
      else if (strcmp((char*)temp_buf, "W") == 0) {
	
	// get the word string and check the lexicon if this word
	// exists
	//
	temp_buf = (char_1*)strtok((char*)NULL, " ");
	hcell = words_a->hash_lookup_cc((char_1*)temp_buf);
	
	// error if word not found in lexicon
	//
	if (hcell == (Train_Hash_cell*)NULL) {
	  fprintf (stdout, "%s\n", temp_buf);
	  error_handler_cc((char_1*)"lat_read_0.cc",
			   (char_1*)"word does not exist in lexicon");
	}	
	
	// set the item index of the node
	//
	word = (Train_Word*)(hcell->get_item_cc());
	lnodes[node_index]->set_word_cc(word);
	read_words = ISIP_FALSE;
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // read in the pronunciation variant used
      //
      else if (strcmp((char*)temp_buf, "v") == 0) {

	// do not confuse with alphabet L which could be a word
	//
	if (read_words == ISIP_TRUE) {

	  // set the pronunciation variation index
	  //
	  temp_buf = (char_1*)strtok((char*)NULL, " ");
	  pron_var = atoi((char*)temp_buf);
	  lnodes[node_index]->set_pron_var_cc(pron_var);
	}

	// otherwise ignore this word and move on
	//
	else {
	  temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	}
      }
      
      // otherwise ignore this word and move on
      //
      else {
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
	read_words = ISIP_TRUE;
      }
    } // end while temp_buf is not null
  } // end while fgets

  // start reading the arc information
  // complete parsing the first line of the arc information
  // loop over the remaining lines of the file
  //
  do {

    // find the delimited value
    //
    temp_buf = (char_1*)strtok((char*)buffer_pos, (char*)delimiter);

    // while the line is not empty
    //
    while (temp_buf != (char_1*)NULL) {
      
      // move the pointer to first non-whitespace
      //
      while (isspace((char)(*temp_buf))) {
	temp_buf++;
      }
      
      // read in the arc index
      //
      if (strcmp((char*)temp_buf, "J") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, " ");
	arc_count++;
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // start node of arc
      //
      else if (strcmp((char*)temp_buf, "S") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, " ");
	start_node = atoi((char*)temp_buf);
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // end node of arc
      //
      else if (strcmp((char*)temp_buf, "E") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, " ");
	end_node = atoi((char*)temp_buf);
	
	// update the prev node and next node pointers for the two nodes
	// corresponding to this arc
	//
	lnodes[start_node]->add_next_node_cc(lnodes[end_node]);
	lnodes[end_node]->add_prev_node_cc(lnodes[start_node]);
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // word represented by the arc end node
      //
      else if (strcmp((char*)temp_buf, "W") == 0) {
	
	// find the corresponding word in the lexicon
	//
	temp_buf = (char_1*)strtok((char*)NULL, " ");
	hcell = words_a->hash_lookup_cc((char_1*)temp_buf);
	
	// error if word not found in lexicon
	//
	if (hcell == (Train_Hash_cell*)NULL) {
	  fprintf (stdout, "%s\n", (char*)temp_buf);
	  error_handler_cc((char_1*)"lat_read_0.cc",
			   (char_1*)"word does not exist in lexicon");
	}	
	
	// set the item index of the node
	//
	word = (Train_Word*)(hcell->get_item_cc());
	lnodes[end_node]->set_word_cc(word);
	
	// create the Train_Lattice start node --- this is a dummy node that is
	// not evaluated, it just serves as a starting point for all the
	// sentence start nodes
	//
	if (strcmp((char*)temp_buf, (char*)TRAIN_LATTICE_SENT_START) == 0) {
	  start_node_d = lnodes[start_node];
	}
	
	// move on to the next word
	//
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // pronunciation variant
      //
      else if (strcmp((char*)temp_buf, "v") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, " ");
	pron_var = atoi((char*)temp_buf);
	lnodes[end_node]->set_pron_var_cc(pron_var);
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // acoustic likelihood of arc
      //
      else if (strcmp((char*)temp_buf, "a") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, " ");
	ac_like = atof((char*)temp_buf);
	lnodes[start_node]->add_ac_score_cc(ac_like);
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // read in the lm score
      //
      else if (strcmp((char*)temp_buf, "l") == 0) {
	temp_buf = (char_1*)strtok((char*)NULL, " ");
	lm_score = atof((char*)temp_buf);
	lm_score *= lm_scale_d;
	lnodes[start_node]->add_lm_score_cc(lm_score);
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
      
      // move on to the next word
      //
      else {
	temp_buf = (char_1*)strtok((char*)NULL, (char*)delimiter);
      }
    }
  } while ((fgets((char*)buffer_pos,ISIP_MAX_STRING_LENGTH,fp_in_a)) &&
	   (arc_count <= num_arcs_d));

  // free memory
  //
  if (lnodes != (Train_Lattice_node**)NULL) {
    delete [] lnodes;
  }
  delete [] buffer;
 
  // exit gracefully
  //
  return ISIP_TRUE;
}