lat_read_1.cc

这是处理语音信号的程序

// file: lat_read_1.cc
//

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

// isip include files
//
#include "lattice.h"
#include "lattice_constants.h"

// method: read_trans_cc
//
// arguments:
//  char_1* trans_str: (input) the transcription for the current mfcc file
//  Hash_table* lexicon_a: (input) the hash table of word indices
//
// return: a logical_1 indicating status
//
// this method reads in the lattice from a file
//
logical_1 Lattice::read_trans_cc(char_1* trans_str_a, Hash_table* lexicon_a) {

  // get the null word from the lexicon
  //
  if ((lexicon_a->hash_lookup_cc(WRD_NULL)) != (Hash_cell*)NULL) {
    null_word_d =
      (Word*)((lexicon_a->hash_lookup_cc(WRD_NULL))->get_item_cc());
  }
  
  // local variables
  //
  int_4 string_length = (int_4)0;

  // find the first non-space character
  //
  char_1* tmp_ptr = trans_str_a;
  while (isspace((char)tmp_ptr[0])) {
    tmp_ptr++;
  }

  // copy the transcription string
  //
  char_1* str = (char_1*)NULL;
  char_1* buffer = new char_1[strlen((char*)tmp_ptr) + 1];
  strcpy((char*)buffer, (char*)tmp_ptr);

  while (isspace(buffer[strlen((char*)buffer) - 1])) {
    buffer[strlen((char*)buffer) - 1] = '\0';
  }

  // dummy variables
  //
  Word** words = new Word*[LATTICE_MAX_NUM_WORD];
  Hash_cell* hcell = (Hash_cell*)NULL;

  // start with the sentence start word
  //
  int_4 num_words = (int_4)0;
  hcell = lexicon_a->hash_lookup_cc(WRD_SENT_START);
  words[num_words++] = (Word*)(hcell->get_item_cc());
  
  // break the word sequence from input string
  //
  str = (char_1*)strtok((char*)buffer, ISIP_STRING_SPACE);
  hcell = lexicon_a->hash_lookup_cc(str);
  if (hcell == (Hash_cell*)NULL) {
    error_handler_cc((char_1*)"read_trans_cc",
                     (char_1*)"word not in lexicon");
  }
  words[num_words++] = (Word*)(hcell->get_item_cc());
  
  // get the rest of the words
  //
  while ((str = (char_1*)strtok((char*)NULL, ISIP_STRING_SPACE)) &&
	 str[0] != ISIP_NEWLINE) {
    
    // replace last character of string with the null string terminator
    // this this is the last word of the transcription
    //
    string_length = strlen((char*)str);
    if (str[string_length-1] == (char_1)'\n') {
      str[string_length-1] = (char_1)'\0';
    }
    hcell = lexicon_a->hash_lookup_cc(str);
    if (hcell == (Hash_cell*)NULL) {
      error_handler_cc((char_1*)"read_trans_cc",
                       (char_1*)"word not in lexicon");
    }
    words[num_words++] = (Word*)(hcell->get_item_cc());
  }

  // stop with the sentence end word
  //
  hcell = lexicon_a->hash_lookup_cc(WRD_SENT_END);
  words[num_words++] = (Word*)(hcell->get_item_cc());

  // free memory
  //
  delete [] buffer;
  
  // set number of nodes and arcs
  //
  num_nodes_d = num_words + 1;
  num_arcs_d = num_words;

  // allocate space for lattice node list
  //
  Lattice_node** lnodes = new Lattice_node*[num_nodes_d];
  float_4 lm_score = (float_4)0;
  float_4 ac_score = (float_4)0;

  // put them in the lattice hash table
  //
  init_nodes_cc(lnodes);

  // create the 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
  //
  start_node_d = lnodes[0];

  // loop over all words i.e. lattice nodes
  //
  for (int_4 sn = 0, en = 1; sn < num_words; sn++, en++) {

    // set the item index of the node
    //
    lnodes[en]->set_word_cc(words[sn]);
    
    // update the prev node and next node pointers for the two nodes
    // corresponding to this arc
    //
    lnodes[sn]->add_next_node_cc(lnodes[en]);
    lnodes[en]->add_prev_node_cc(lnodes[sn]);
    
    // set the acoustic and language model scores
    //
    lnodes[sn]->add_lm_score_cc(lm_score);
    lnodes[sn]->add_ac_score_cc(ac_score);
  }

  // free memory
  //
  delete [] words;
  delete [] lnodes;
  
  // exit gracefully
  //
  return ISIP_TRUE;
}