ghmm_emission.cpp

General Hidden Markov Model Library 一个通用的隐马尔科夫模型的C代码库

/*
 * created: 21 Feb 2002 by Peter Pipenbacher
 * authors: Peter Pipenbacher ([e-mail])
 * file   : $Source$
 * $Id: GHMM_Emission.cpp 275 2003-09-18 10:04:49Z cic99 $
 *
 * __copyright__
 */

#include <xmlio/XMLIO_Document.h>
#include "ghmm++/GHMM_Emission.h"
#include "ghmm++/GHMM_State.h"
#include "ghmm++/GHMM_ContinuousModel.h"
#include "ghmm++/GHMM_DiscreteModel.h"
#include "ghmm++/GHMM_Alphabet.h"

#include <iostream>

#ifdef HAVE_NAMESPACES
using namespace std;
#endif


XMLIO_Element* GHMM_CEmission::XMLIO_startTag(const string& tag, XMLIO_Attributes &attrs) {
  bool found = false; 
  if (tag == "gauss") {
    mue.push_back(atof(attrs["mue"].c_str()));
    variance.push_back(atof(attrs["variance"].c_str()));
    density  = normal;
    found    = true;
  }
  if (tag == "gauss-positive") {
    mue.push_back(atof(attrs["mue"].c_str()));
    variance.push_back(atof(attrs["variance"].c_str()));
    density  = normal_pos;
    found    = true;
  }
  if (tag == "gauss-approximated") {
    mue.push_back(atof(attrs["mue"].c_str()));
    variance.push_back(atof(attrs["variance"].c_str()));
    density  = normal_approx;
    found    = true;
  }
  
  if (! found) {
    fprintf(stderr,"<emission> element of state '%s' has unrecognized tag '%s'.\n",
	    state->id.c_str(),tag.c_str());
    exit(1);
  }
  
  if (attrs["mue"] == "") {
    fprintf(stderr,"<%s> element of state '%s' lacks mue attribute.\n",
	    tag.c_str(),state->id.c_str());
    exit(1);
  }

  if (attrs["variance"] == "") {
    fprintf(stderr,"<%s> element of state '%s' lacks variance attribute.\n",
	    tag.c_str(),state->id.c_str());
    exit(1);
  }

  return NULL;
}


void GHMM_CEmission::XMLIO_getCharacters(const string& characters) {
  unsigned int pos;
  for (pos = 0; pos < characters.size(); ++pos) {
    while (pos < characters.size() && isspace(characters[pos]))
      ++pos;

    if (pos < characters.size())
      weights.push_back(atof(characters.substr(pos).c_str()));

    while (pos < characters.size() && !isspace(characters[pos]))
      ++pos;
  }
}

void GHMM_CEmission::XMLIO_finishedReading() {
  /* continuous model */
  if (weights.size() != mue.size()) {
    fprintf(stderr,"Different number of weights and density functions in state '%s'.\n",state->id.c_str());
    exit(1);
  }
}

const int GHMM_CEmission::XMLIO_writeContent(XMLIO_Document& writer) {
  int result = 0;
  writer.changeIndent(2);
  
  /* continuous model */
  if (state->c_sstate) {
    result = writer.writef("1 <");
    switch (density) {
    case normal:
      result += writer.write("gauss");
      break;
    case normal_pos:
      result += writer.write("gauss-positive");
      break;
    case normal_approx:
      result += writer.write("gauss-approximated");
      break;
    default:
      break;
    }
    result += writer.writef(" mue=\"%f\" variance=\"%f\">",mue[0],variance[0]);
  }
  return result;
}

const int GHMM_DEmission::XMLIO_writeContent(XMLIO_Document& writer) {
  int result = 0;
  int i;
  
  writer.changeIndent(2);

  GHMM_Alphabet* alphabet = state->getModel()->getAlphabet();
  GHMM_DiscreteModel* model = (GHMM_DiscreteModel*) state->getModel();

  result += writer.writeEndl();
  for (i = 0; i < model->c_model->M; ++i) {
    result += writer.writef("%s%.2f",writer.indent,state->c_state->b[i]);
    if (alphabet)
      result += writer.writef(" <!-- %s -->",alphabet->getSymbol(i).c_str());
    result += writer.writeEndl();
  }

  return result;
}