tp_comp_0.cc

这是处理语音信号的程序

// file: tp_comp_0.cc
//

// isip include files
//
#include "trace_projector.h"
#include "trace_projector_constants.h"
#include <time.h>

// method: compute_cc
//
// arguments: none
//
// return: a logical flag to indicate success
//
logical_1 Trace_projector::compute_cc() {

  // if this is lattice error rate mode, only need the lattice and
  // transcription files
  //
  if (function_mode_d == TP_LATTICE_WER_FUNCTION) {

    // read the lattice list file
    //
    if (lattice_file_d == (char_1*)NULL) {
      fprintf(stdout, "Error: input lattice list file not specified\n");
      exit(ISIP_PROTO_ERROR);
    }
    
    // open lattice file list
    //
    FILE* flat = (FILE*)NULL;
    open_file_cc(flat, lattice_file_d, (char_1*)"r");

    // read the utterance list file
    //
    if (forcealign_file_d == (char_1*)NULL) {
      fprintf(stdout, "Error: forced alignment file not specified\n");
      exit(ISIP_PROTO_ERROR);
    }
    
    // open transcription i.e. forced alignment file
    //
    FILE* fref = (FILE*)NULL;
    open_file_cc(fref, forcealign_file_d, (char_1*)"r");
    
    // open the output file
    //
    FILE* fout = (FILE*)NULL;
    open_file_cc(fout, latwer_file_d, (char_1*)"w");

    // compute the lattice wer
    //
    fprintf(stdout, "Computing lattice word error rates...\n");
    decoder_d->lattice_wer_cc(flat, fref, fout);

    // close files
    //
    if (flat != (FILE*)NULL) {
      fclose(flat);
    }
    if (fref != (FILE*)NULL) {
      fclose(fref);
    }
    if (fout != (FILE*)NULL) {
      fclose(fout);
    }
  }

  // otherwise this is decoding mode
  //
  else {
  
    // load acoustic model related data into the decoder and clean
    // related memory
    //
    fprintf(stdout, "Loading acoustic model data...\n");
    load_ac_models_cc();
    
    // if ngram required in the specified function mode, load the ngram
    // language model
    //
    if ((function_mode_d == TP_NGRAM_DECODE_FUNCTION) ||
	(function_mode_d == TP_LATTICE_GENERATE_FUNCTION) ||
	(function_mode_d == TP_LATTICE_VERIFY_FUNCTION)) {
      
      FILE* fp;

      // check if the ngram lm file exists
      //
      if (lm_file_d == (char_1*)NULL) {
	fprintf(stdout, "Error: language model file not specified\n");
	exit(ISIP_PROTO_ERROR);
      }
      open_file_cc(fp, lm_file_d, (char_1*)"r");
      
      // load the ngram
      //
      fprintf(stdout, "Reading language model data...\n");
      decoder_d->read_ngram_cc(fp);

      if (fp != (FILE*)NULL) {
	fclose(fp);
      }

      // check if the list of grammars is specified
      //
      if (gram_list_file_d != (char_1*)NULL) {

	// open file
	//
	open_file_cc(fp, gram_list_file_d, (char_1*)"r");

	// load the grammars for hierarchical grammar processing
	//
	decoder_d->read_gram_list_cc(fp);

	if (fp != (FILE*)NULL) {
	  fclose(fp);
	}
      }
    }

    // set the user specified lm scaling and word insertion values  
    //
    decoder_d->set_lmscale_cc(lm_scale_d);
    decoder_d->set_wdpenalty_cc(word_penalty_d);
    
    // set the acoustic context mode
    //
    decoder_d->set_context_mode_cc(context_mode_d);
    
    // set the pruning related parameters and clean related memory 
    //
    decoder_d->set_beams_cc(beam_width_d);
    decoder_d->set_mapmi_limit_cc(mapmi_limit_d);
    decoder_d->set_max_words_cc(num_wdcut_d);
    
    // set the nbest pruning related parameters
    //
    decoder_d->set_nbest_maxpaths_cc(nbest_max_paths_d);
    decoder_d->set_nbest_beam_cc(nbest_beam_d);

    // initialize the decoder for hypothesizing utterances
    //
    decoder_d->init_cc();

    // file pointers for processing lists of files
    //
    FILE* fpin_list = (FILE*)NULL;
    FILE* fpout_list = (FILE*)NULL;
    FILE* fpli_list = (FILE*)NULL;
    FILE* fplo_list = (FILE*)NULL;
    FILE* fpnbo_list = (FILE*)NULL;
    FILE* fpms_list = (FILE*)NULL;
    
    // open the file containing the list of input mfcc files
    //
    if (input_file_d == (char_1*)NULL) {
      fprintf(stdout, "Error: input mfcc file list not specified\n");
      exit(ISIP_PROTO_ERROR);
    }
    open_file_cc(fpin_list, input_file_d, (char_1*)"r");

    // open the file containing the list of output files
    //
    if (output_file_d == (char_1*)NULL) {
      fprintf(stdout, "Error: output file list not specified\n");
      exit(ISIP_PROTO_ERROR);
    }
    open_file_cc(fpout_list, output_file_d, (char_1*)"r");

    // open the file containing the list of model-level segment files
    //
    if (segment_mode_d == TP_MODEL_SEGMENT_MODE) {
      
      // check if file exists
      //
      if (model_segment_file_d == (char_1*)NULL) {
	fprintf(stdout, "Error: input model-level segment list file not specified\n");
	exit(ISIP_PROTO_ERROR);
      }
      
      // open model-level segment file list
      //
      open_file_cc(fpms_list, model_segment_file_d, (char_1*)"r");
    }
    
    // open the file containing the list of lattice files
    //
    if ((function_mode_d == TP_LATTICE_RESCORE_FUNCTION) ||
	(function_mode_d == TP_LATTICE_LATTICE_FUNCTION) ||
	(function_mode_d == TP_LATTICE_VERIFY_FUNCTION)) {
      
      // check if file exists
      //
      if (lattice_file_d == (char_1*)NULL) {
	fprintf(stdout, "Error: input lattice list file not specified\n");
	exit(ISIP_PROTO_ERROR);
      }
      
      // open lattice file list
      //
      open_file_cc(fpli_list, lattice_file_d, (char_1*)"r");
    }

    // if this is lattice generate mode there is a list of output
    // lattices
    //
    if ((function_mode_d == TP_LATTICE_GENERATE_FUNCTION) ||
	(function_mode_d == TP_LATTICE_LATTICE_FUNCTION)) {
      
      // check if file exists
      //
      if (outlat_file_d == (char_1*)NULL) {
	fprintf(stdout, "Error: output lattice list file not specified\n");
	exit(ISIP_PROTO_ERROR);
      }
      
      // open lattice file
      //
      open_file_cc(fplo_list, outlat_file_d, (char_1*)"r");
    }

    // open the output nbest list file
    //
    if (((function_mode_d == TP_LATTICE_RESCORE_FUNCTION) ||
	 (function_mode_d == TP_NGRAM_DECODE_FUNCTION) ||
	 (function_mode_d == TP_LATTICE_VERIFY_FUNCTION)) &&
	(num_nbest_d > 1)) {
      
      // check if file exists
      //
      if (outnbest_file_d == (char_1*)NULL) {
	fprintf(stdout, "Error: output nbest list file not specified\n");
	exit(ISIP_PROTO_ERROR);
      }
      
      // open lattice file
      //
      open_file_cc(fpnbo_list, outnbest_file_d, (char_1*)"r");
    }
    
    // open the file containing reference transcriptions for forced
    // alignment
    //
    if (function_mode_d == TP_FORCE_ALIGN_FUNCTION) {
      
      // check if file exists
      //
      if (forcealign_file_d == (char_1*)NULL) {
	fprintf(stdout, "Error: forced alignment file not specified\n");
	exit(ISIP_PROTO_ERROR);
      }
      
      // open forced alignment file
      //
      open_file_cc(fpli_list, forcealign_file_d, (char_1*)"r");
    }

    // define file pointers for data files
    //
    FILE* fpi = (FILE*)NULL;
    FILE* fpo = (FILE*)NULL;
    FILE* fpli = (FILE*)NULL;
    FILE* fplo = (FILE*)NULL;
    FILE* fpnbo = (FILE*)NULL;
    FILE* fpmsi = (FILE*)NULL;
    
    // allocate memory for individual mfcc, lattice and output files
    //
    char_1* feat_file = new char_1[ISIP_MAX_STRING_LENGTH];
    char_1* hyp_file = new char_1[ISIP_MAX_STRING_LENGTH];
    char_1* lattice_in = new char_1[ISIP_MAX_STRING_LENGTH];
    char_1* lattice_out = new char_1[ISIP_MAX_STRING_LENGTH];
    char_1* transcription = new char_1[ISIP_MAX_STRING_LENGTH];
    char_1* nbest_out = new char_1[ISIP_MAX_STRING_LENGTH];
    char_1* model_seg_in = new char_1[ISIP_MAX_STRING_LENGTH];

    // timiming information variable
    //
    float_8 dec_time = 0, tmp_dec_time;
    int_4 frame = 0, utterance = 0, tmp_frame;

    int_4 num_file = 1;
    
    // read the corresoponding data file from each list and process one
    // by one
    //
    while (fgets((char*)feat_file, ISIP_MAX_STRING_LENGTH, fpin_list) !=
	   (char*)NULL) {

      // skip the new line
      //    
      while (*feat_file == ISIP_NEWLINE) {
	while (fgets((char*)feat_file, ISIP_MAX_STRING_LENGTH, fpin_list) !=
	       (char*)NULL);
	  }
      
      // expand the input filename
      //
      Decoder::expand_filename_cc(feat_file);
      if (input_format_d == TP_ASCII_FORMAT) {
	open_file_cc(fpi, feat_file, (char_1*)"r");
      }
      else {
	open_file_cc(fpi, feat_file, (char_1*)"rb");
      }
      
      // print status message
      //
      fprintf(stdout, "processing file %ld: %s\n", num_file, feat_file);
      fflush(stdout);
      num_file++;