isip_segment_concat.cc

这是一个从音频信号里提取特征参量的程序

// file: $isip/util/speech/isip_segment_concat/isip_segment_concat.cc
// version: $Id: isip_segment_concat.cc,v 1.4 2003/05/07 15:55:11 parihar Exp $
//
// isip include files
//
#include "isip_segment_concat.h"

// isip_concat_segment: utility that extractes the segmental features.
//
int main(int argc, const char** argv) {

  // configure the command line parser and an Sdb object.
  //
  Sdb sdb;
  CommandLine cmdl(sdb);
  
  cmdl.setUsage(
#include "usage_message.text"
                );
  cmdl.setHelp(
#include "help_message.text"
               );

  cmdl.setIdent("$Revision: 1.4 $",
		"$Name: isip_r00_n11 $",
		"$Date: 2003/05/07 15:55:11 $");
  
  // add a command line flag for the min-max file generation
  //
  Boolean min_max_flag;
  cmdl.addFlagParam(min_max_flag, OPTION_MIN_MAX_FLAG);

  // add a command line option for the min-max (input/output) file type
  //
  String min_max_file_type;
  cmdl.addOptionParam(min_max_file_type, OPTION_MIN_MAX_TYPE, DEF_FILE_TYPE);

  // add a command line option for the min-max (input/output) file
  //
  Filename min_max_file;
  cmdl.addOptionParam(min_max_file, OPTION_MIN_MAX_FILE, (Filename)EMPTY); 

  // add a command line option for the input feature vector dimension
  // incase the features are in raw format.
  //
  Long dim(0);
  cmdl.addOptionParam(dim, OPTION_DIM, DEF_DIM);

  // add a command line option for the format of the input feature-files
  //
  String input_file_format;
  cmdl.addOptionParam(input_file_format, OPTION_INPUT_FILE_FORMAT,
		      DEF_FILE_FORMAT);
  
  // add a command line option for the type of the input feature-files
  //
  String input_file_type;
  cmdl.addOptionParam(input_file_type, OPTION_INPUT_FILE_TYPE, DEF_FILE_TYPE);
  
  // add a command line option for the debug level
  //
  DebugLevel debug_level;
  cmdl.addOptionParam(debug_level, OPTION_DEBUG_LEVEL, DEBUG_LEVEL_DEFAULT);

  // add a command line option for the log file
  //
  Filename log_file;
  cmdl.addOptionParam(log_file, OPTION_LOG_FILE, (Filename)LOG_FILE_DEFAULT);

  // add a command line option for output file type
  //
  String output_file_type;
  cmdl.addOptionParam(output_file_type, OPTION_OUTPUT_FILE_TYPE, EMPTY);

  // add a command line option for the format of the output feature-files
  //
  String output_file_format;
  cmdl.addOptionParam(output_file_format, OPTION_OUTPUT_FILE_FORMAT,
		      DEF_FILE_FORMAT);
  
  // add a command line option for suffix
  //
  String output_suffix;
  cmdl.addOptionParam(output_suffix, OPTION_OUTPUT_SUFFIX, EMPTY);      

  // add a command line option for ratio
  //
  String ratio;
  cmdl.addOptionParam(ratio, OPTION_RATIO, EMPTY);      

  // add a command line option for transcription database
  //
  Filename transdb_file;
  cmdl.addOptionParam(transdb_file, OPTION_TRANS_DB, (Filename)EMPTY);      

  // add a command line option for audio database
  //
  Filename audiodb_file;
  cmdl.addOptionParam(audiodb_file, OPTION_AUDIO_DB, (Filename)EMPTY);      

  // add a command line option for transcription level in the
  // transcription database
  //
  String trans_level;
  cmdl.addOptionParam(trans_level, OPTION_TRANS_LEVEL, EMPTY);      

  // add a command line flag for log-duration feature
  //
  Boolean duration_flag;
  cmdl.addFlagParam(duration_flag, OPTION_DURATION_FLAG);

  // add a command line flag for the normalization of the mfcc
  // features before extracting the segmental features
  //
  Boolean normalize_flag;
  cmdl.addFlagParam(normalize_flag, OPTION_NORMALIZE_FLAG);

  // parse the command line
  //
  if (!cmdl.parse(argc, argv)) {
    cmdl.printUsage();
  }
  
  // open the log file
  //
  if (log_file.ne(L"-")) {
    Console::open(log_file);
    Console::setLineWrap(File::NO_WRAP);
  }

  // case 1: if min_max flag is set, compute the min and max in each
  // dimension
  //
  if (min_max_flag) {
    
    // verification incase the user specifies the min_max_flag, check
    // for min_max_file
    //
    if (min_max_file.eq((Filename)EMPTY)) {
      cmdl.printUsage();
    }
    
    // verification for the audio database file
    //
    if (audiodb_file.eq((Filename)EMPTY)) {
      cmdl.printUsage();
    }
    
    // verification incase the user specifies the feature-file format as
    // RAW, then, we need the dimension
    //
    if (input_file_format.eq(FILE_FORMAT_RAW) && ((long)dim <= (long)0)) {
      cmdl.printUsage();
    };
    
    // set the file format
    //
    FeatureFile::FILE_FORMAT in_f_format = FeatureFile::SOF;
    if (input_file_format.eq(FILE_FORMAT_RAW)) {
      in_f_format = FeatureFile::RAW;
    }
  
    // set the file type
    //
    FeatureFile::FILE_TYPE in_f_type = FeatureFile::TEXT;
    if (input_file_type.eq(FILE_TYPE_BINARY)) {
      in_f_type = FeatureFile::BINARY;
    }
    
    // compute the min-max values
    //
    Vector<VectorFloat> min_max;
    SegmentConcat sc;
    sc.setDebug(debug_level);
    sc.setDimension(dim);
    sc.compute(sdb, audiodb_file, in_f_type, in_f_format, min_max);
    
    // write the min-max values to the file
    //
    Sof tmp_sof;
    VectorFloat temp;
    tmp_sof.open(min_max_file, File::WRITE_ONLY);
    min_max.write(tmp_sof, (long)0);
    tmp_sof.close();
  }
  
  // case 2: compute the segmental features
  //
  else  {    
    
    // verification incase the user specifies the feature-file format as
    // RAW, then, we need the dimension
    //
    if (input_file_format.eq(FILE_FORMAT_RAW) && ((long)dim <= (long)0)) {
      cmdl.printUsage();
    };
    
    // verification for ratio
    //
    if (ratio.eq(EMPTY)) {
      cmdl.printUsage();
    };
    
    // verification for the audio database file
    //
    if (audiodb_file.eq((Filename)EMPTY)) {
      cmdl.printUsage();
    }
    
    // verification for transcription database
    //
    if (transdb_file.eq((Filename)EMPTY)) {
      cmdl.printUsage();
    };

    // verification for transcription level
    //
    if (trans_level.eq(EMPTY)) {
      cmdl.printUsage();
    };

    // specifics for normalize flag
    //
    if (normalize_flag) {
      
      // verification incase the user specifies the normalize flag,
      // check for min_max_file
      //
      if (min_max_file.eq((Filename)EMPTY)) {
	cmdl.printUsage();
      }
    }

    // set the file format
    //
    FeatureFile::FILE_FORMAT in_f_format = FeatureFile::SOF;
    if (input_file_format.eq(FILE_FORMAT_RAW)) {
      in_f_format = FeatureFile::RAW;
    }
    FeatureFile::FILE_FORMAT o_f_format = FeatureFile::SOF;
    if (output_file_format.eq(FILE_FORMAT_RAW)) {
      o_f_format = FeatureFile::RAW;
    }
  
    // set the file type
    //
    FeatureFile::FILE_TYPE in_f_type = FeatureFile::TEXT;
    if (input_file_type.eq(FILE_TYPE_BINARY)) {
      in_f_type = FeatureFile::BINARY;
    }
    FeatureFile::FILE_TYPE o_f_type = FeatureFile::TEXT;
    if (output_file_type.eq(FILE_TYPE_BINARY)) {
      o_f_type = FeatureFile::BINARY;
    }

    // compute the segmental features
    //
    SegmentConcat sc;
    sc.setDebug(debug_level);
    sc.setDimension(dim);
    sc.compute(sdb, audiodb_file, transdb_file, in_f_type, in_f_format,
	       trans_level, ratio, duration_flag, normalize_flag, min_max_file,
	       o_f_type, o_f_format, output_suffix);

  }
  
  // exit gracefully
  //
  return Integral::exit();
}