FusionScore.cpp

Go to the documentation of this file.

/*
This file is part of LIA_RAL which is a set of software based on ALIZE
toolkit for speaker recognition. ALIZE toolkit is required to use LIA_RAL.

LIA_RAL project is a development project was initiated by the computer
science laboratory of Avignon / France (Laboratoire Informatique d'Avignon -
LIA) [http://lia.univ-avignon.fr <http://lia.univ-avignon.fr/>]. Then it
was supported by two national projects of the French Research Ministry:
        - TECHNOLANGUE program [http://www.technolangue.net]
        - MISTRAL program [http://mistral.univ-avignon.fr]

LIA_RAL is free software: you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as
published by the Free Software Foundation, either version 3 of
the License, or any later version.

LIA_RAL is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public
License along with LIA_RAL.
If not, see [http://www.gnu.org/licenses/].

The LIA team as well as the LIA_RAL project team wants to highlight the
limits of voice authentication in a forensic context.
The "Person Authentification by Voice: A Need of Caution" paper
proposes a good overview of this point (cf. "Person
Authentification by Voice: A Need of Caution", Bonastre J.F.,
Bimbot F., Boe L.J., Campbell J.P., Douglas D.A., Magrin-
chagnolleau I., Eurospeech 2003, Genova].
The conclusion of the paper of the paper is proposed bellow:
[Currently, it is not possible to completely determine whether the
similarity between two recordings is due to the speaker or to other
factors, especially when: (a) the speaker does not cooperate, (b) there
is no control over recording equipment, (c) recording conditions are not
known, (d) one does not know whether the voice was disguised and, to a
lesser extent, (e) the linguistic content of the message is not
controlled. Caution and judgment must be exercised when applying speaker
recognition techniques, whether human or automatic, to account for these
uncontrolled factors. Under more constrained or calibrated situations,
or as an aid for investigative purposes, judicious application of these
techniques may be suitable, provided they are not considered as infallible.
At the present time, there is no scientific process that enables one to
uniquely characterize a persones voice or to identify with absolute
certainty an individual from his or her voice.]

Copyright (C) 2004-2010
Laboratoire d'informatique d'Avignon [http://lia.univ-avignon.fr]
LIA_RAL admin [alize@univ-avignon.fr]
Jean-Francois Bonastre [jean-francois.bonastre@univ-avignon.fr]
*/

#if !defined(ALIZE_FusionScore_cpp)
#define ALIZE_FusionScore_cpp

#include <iostream>
#include <fstream>  // pour outFile
#include <cstdio>   // pour printf()
#include <cassert> // pour le debug pratique
#include <cmath>
#include "FusionScore.h"
#include <liatools.h>

using namespace alize;
using namespace std;

//-------------------------------------------------------------------------
void computeFusion(XLine & lineTab, XLine & finalScore, XList & weights,Histo &destH, Config & config)
{
        String fusionMethod=config.getParam("fusionMethod");

        if (fusionMethod=="ArithMean")
        {
                double Score=0;
                double sumWeight=0;
                for (unsigned long i=0; i<lineTab.getElementCount(); i++)
                {
                        cout << weights.getLine(0).getElement(i) << " * " << lineTab.getElement(i) << endl;
                        sumWeight += weights.getLine(0).getElement(i).toDouble();
                        Score+=weights.getLine(0).getElement(i).toDouble()*lineTab.getElement(i).toDouble();
                }
                Score/=sumWeight; 
                finalScore.addElement(String::valueOf(Score));
        }
        else if (fusionMethod=="GeoMean")
        {
                double Score=1;
                double sumWeight=0;
                double nbElm=lineTab.getElementCount();
                for (int i=0; i<nbElm; i++){
                        sumWeight += weights.getLine(0).getElement(i).toDouble();                       
                        Score*=pow(weights.getLine(0).getElement(i).toDouble(),lineTab.getElement(i).toDouble());
                }
                if (Score < 0)  {Score=pow(-1*Score,1.0/sumWeight); Score=-1*Score;}
                else {Score=pow(Score,1.0/sumWeight);}
                finalScore.addElement(String::valueOf(Score));
        }
        else if (fusionMethod=="multipleAryth"){
                double Score=0;
                // ici provoque warning  de compilation convert from double to int
                // long bin= destH(lineTab.getElement(0).toDouble(),1); 
                // modifié par éric -> introduction de cast to long pour supprimer warning
          long bin= (long)destH(lineTab.getElement(0).toDouble(),1); 

                if (debug) cout <<"score["<<lineTab.getElement(0).toDouble()<<"], fusion with line["<<bin<<"]"<<endl;
                for (unsigned long i=0; i<lineTab.getElementCount(); i++)
                {
                        Score+=weights.getLine(bin).getElement(i).toDouble()*lineTab.getElement(i).toDouble();
                }
                //Score/=lineTab.getElementCount(); give percentage in weight, it's better
                finalScore.addElement(String::valueOf(Score));
        }

        else {cerr <<"fusionMethod"<<fusionMethod<<"] unknown"<<endl; exit(1);}
}

//-------------------------------------------------------------------------
void findNbest(XLine & tabScoreLine, int Nbest, int comp_double(const void *a, const void *b))
{
        int nbElm=tabScoreLine.getElementCount();
        double *tab = new double[nbElm];

        for (int i=0; i < nbElm; i++)
                tab[i]=tabScoreLine.getElement(i).toDouble();

        qsort(tab,nbElm,sizeof(double),comp_double);

        tabScoreLine.reset();

        for (int j=nbElm-1; j >= nbElm-Nbest; j--)
                tabScoreLine.addElement(String::valueOf(tab[j]));
}

//-------------------------------------------------------------------------
int comp_double(const void *a, const void *b)
{
        if ((*(double*) a - *(double*) b)<=0)   {return -1;}
        else    {return 1;}
}

String extractModel(String &classModel){
        String tmp;
        String m = classModel;
        long ind =  m.find("_", 0);
        for(int i=0; i<ind; i++) {
                tmp += m[i];
        }
        return tmp;
}

//-------------------------------------------------------------------------
int FusionScore(Config& config)
{
        using namespace alize;
        using namespace std;

        String inputNISTListFileName = config.getParam("inputFileList");
        String outputNISTFileName = config.getParam("outputFile");
        String weightsFile = config.getParam("weights");
        String format=config.getParam("format");
        bool debug=false;
        if (config.existsParam("debug")) debug=true;
        String fusionMethod=config.getParam("fusionMethod");
        Histo destH;
        if (fusionMethod=="multipleAryth")
          destH.load(config.getParam("destH"));
        unsigned long nbFiles;
        unsigned long genderField, nameField, decisionField, segField, scoreField;              //Necessary field in files
        if (format=="lia") {setLIAInfoFields(genderField, nameField, decisionField, segField,scoreField);}
        else if(format=="nist") {setNIST04InfoFields(genderField, nameField, decisionField, segField,scoreField);}
        else if (format=="etf") {scoreField=7;} // To remove, but sorry we're in eval!! retrieveETFInfoFields(genderField, nameField, decisionField, segField,scoreField);}
        else {cerr << "Error: Format unknown" << endl; exit(1);}
        
        XList tabScore;
        XLine finalScore;

        try{
                XList inputNISTList(inputNISTListFileName,config);      // List of files to fuse
                XList weights(weightsFile,config);      // coeffs
                nbFiles=inputNISTList.getLineCount();

                for (unsigned long i=0; i< nbFiles; i++) {
                        XList inputNISTFile(inputNISTList.getLine(i).getElement(0),config);
                        for (unsigned long j=0; j < inputNISTFile.getLineCount(); j++) {
                                if (i==0) {tabScore.addLine();}
                                tabScore.getLine(j).addElement(inputNISTFile.getLine(j).getElement(scoreField));
                        }
                }

                for (unsigned long i=0; i < tabScore.getLineCount(); i++)       {//Fusion Process
                        if (config.existsParam("Nbest"))        { // In case of a nbest fusion
                                findNbest(tabScore.getLine(i), config.getParam("Nbest").toLong(), comp_double);
                        }

                if (debug) {
                        for (unsigned long j=0; j < tabScore.getLine(i).getElementCount(); j++)
                                cout << tabScore.getLine(i).getElement(j) << " ";
                                cout << endl;
                }
                        computeFusion(tabScore.getLine(i),finalScore,weights,destH,config); // See above method
                }
                XList inputNISTFile(inputNISTList.getLine(0).getElement(0),config);
                ofstream outFile(outputNISTFileName.c_str(),ios::out | ios::trunc);     // Preparing output
                String subType, event, channel, source;
                unsigned long duration, start;
                double LLR;
                int decision=0; // no decision in the fusion process
 
                if (format=="lia") {
                        for (unsigned long i=0; i < tabScore.getLineCount(); i++) {
                                setLIAInfoFields(genderField, nameField, decisionField, segField, scoreField); // retrieve LIA fields
                                subType=inputNISTFile.getLine(i).getElement(genderField);
                                event=inputNISTFile.getLine(i).getElement(nameField);
                                decision=decision; // there is a pb here, outputResult takes a int as decision
                                source=inputNISTFile.getLine(i).getElement(segField);
                                LLR=finalScore.getElement(i).toDouble();
                                outputResultLine(LLR,event,source, subType,decision,outFile);   
                        }
                }
                if (format=="nist") {
                        for (unsigned long i=0; i < tabScore.getLineCount(); i++) {
                                String trainTypeTest="1side"; // clean this ASAP
                                String adaptationMode="n";
                                String segTypeTest="1side";
                                setNIST04InfoFields(genderField, nameField, decisionField, segField, scoreField); // retrieve NIST fields
                                subType=inputNISTFile.getLine(i).getElement(genderField);
                                event=inputNISTFile.getLine(i).getElement(nameField);
                                String decision="-";// there is a pb here, outputResult takes a int as decision
                                source=inputNISTFile.getLine(i).getElement(segField);
                                LLR=finalScore.getElement(i).toDouble();
                                outputResultNIST04Line(trainTypeTest,adaptationMode,segTypeTest,subType,event,source,decision,LLR,outFile);                             
                        }
                }
                else if (format=="etf") {
                        for (unsigned long i=0; i < tabScore.getLineCount(); i++) { // retrieve files from etf format has to be done
                                subType=inputNISTFile.getLine(i).getElement(0);
                                event=inputNISTFile.getLine(i).getElement(1);
                                decision=inputNISTFile.getLine(i).getElement(2).toLong();
                                source=inputNISTFile.getLine(i).getElement(3);
                                start=inputNISTFile.getLine(i).getElement(4).toLong();
                                duration=inputNISTFile.getLine(i).getElement(5).toLong();
                                LLR=finalScore.getElement(i).toDouble();
                                outputResultLine(LLR,event,source, start, start+duration, subType,decision,outFile);    
                        }       
                }       
} // fin try

catch (Exception& e)
        { 
          cout <<" FusionScore" << e.toString().c_str() << endl;
        }
        return 0;
}


int FusionClassScore(Config& config)
{

        using namespace alize;
        using namespace std;

        String inputNISTFileName = config.getParam("inputNistFileName");
        String outputNISTFileName = config.getParam("outputFile");
        ofstream outFile(outputNISTFileName.c_str(),ios::out | ios::trunc);     // Preparing output
        String format=config.getParam("format");
        bool debug=false;
        if (config.existsParam("debug")) debug=true;

        String fusionMethod=config.getParam("fusionMethod");
        Histo destH;
        if (fusionMethod=="multipleAryth")
          destH.load(config.getParam("destH"));
        
        unsigned long genderField, nameField, decisionField, segField, scoreField;              //Necessary field in files
        if (format=="lia") {setLIAInfoFields(genderField, nameField, decisionField, segField,scoreField);}
        else if(format=="nist") {setNIST04InfoFields(genderField, nameField, decisionField, segField,scoreField);}
        else {cerr << "Error: Format unknown" << endl; exit(1);}

        /* Weight file managing */
        String weightFileExt = ".txt";
        if(config.existsParam("weightFileExt")){
                weightFileExt = config.getParam("weightFileExt");       
        }
        String weightFilePath = "./";
        if(config.existsParam("weightFilePath")){
                weightFilePath = config.getParam("weightFilePath");     
        }
        

        try{
                String subType, event, channel, source;
                double LLR;
                //int decision=0; // no decision in the fusion process
                String trainTypeTest="1side"; // clean this ASAP
                String adaptationMode="n";
                String segTypeTest="1side";

                XList tabScore;
                tabScore.addLine();
                XLine finalScore;
                XList weights;
                weights.addLine();
                XList tmpW;
                
                XList nistList(inputNISTFileName, config);
                XLine * currentlinep = nistList.getLine();
                XLine * nextlinep;
                
                String currentModel = extractModel(currentlinep->getElement(nameField));
                if(debug)
                        cout << "Model: " << currentModel << " Accumule:" << currentlinep->getElement(scoreField) << endl;
                tabScore.getLine(0).addElement(currentlinep->getElement(scoreField));
                String weightFileName = weightFilePath+currentlinep->getElement(nameField)+weightFileExt;
                tmpW.load(weightFileName, config);
                weights.getLine(0).addElement(tmpW.getLine(0).getElement(0));

                while ((nextlinep=nistList.getLine()) != NULL){
                        String nextModel = extractModel(nextlinep->getElement(nameField));              
                        if(nextModel != currentModel){
                                computeFusion(tabScore.getLine(0),finalScore,weights,destH, config); // See above method
                                if(debug)
                                        cout << "Score Fusion: " << finalScore.getElement(0)<< endl;
                                /* Output scores in files */
                                if (format=="lia") {
                                        LLR=finalScore.getElement(0).toDouble();
                                        outputResultLine(LLR,currentModel,currentlinep->getElement(segField),currentlinep->getElement(genderField),int(currentlinep->getElement(decisionField).toLong()),outFile);      
                                }
                                if (format=="nist") {
                                        LLR=finalScore.getElement(0).toDouble();
                                        outputResultNIST04Line(trainTypeTest,adaptationMode,segTypeTest,currentlinep->getElement(genderField),currentlinep->getElement(nameField),currentlinep->getElement(segField),currentlinep->getElement(decisionField),LLR,outFile);                              
                                }
                                
                                currentModel = nextModel;
                                tabScore.getLine(0).reset();
                                weights.getLine(0).reset();
                                finalScore.reset();
                                currentlinep = nextlinep;       
                                if(debug)
                                        cout << "Model: " << currentModel << " Accumule:" << currentlinep->getElement(scoreField) << endl;
                                tabScore.getLine(0).addElement(currentlinep->getElement(scoreField));
                                String weightFileName = weightFilePath+currentlinep->getElement(nameField)+weightFileExt;
                                tmpW.load(weightFileName, config);
                                weights.getLine(0).addElement(tmpW.getLine(0).getElement(0));
                        }
                        else{
                                currentlinep = nextlinep;       
                                if(debug)
                                        cout << "Model: " << currentModel << " Accumule:" << currentlinep->getElement(scoreField) << endl;
                                tabScore.getLine(0).addElement(currentlinep->getElement(scoreField));
                                String weightFileName = weightFilePath+currentlinep->getElement(nameField)+weightFileExt;
                                tmpW.load(weightFileName, config);
                                weights.getLine(0).addElement(tmpW.getLine(0).getElement(0));
                        }
                        
                }
                
                /* last line ???!!!! */
                computeFusion(tabScore.getLine(0),finalScore,weights,destH,config); // See above method
                if(debug)
                        cout << "Score Fusion: " << finalScore.getElement(0)<< endl;
                /* Output scores in files */
                if (format=="lia") {
                        LLR=finalScore.getElement(0).toDouble();
                        outputResultLine(LLR,currentModel,currentlinep->getElement(segField),currentlinep->getElement(genderField),int(currentlinep->getElement(decisionField).toLong()),outFile);      
                }
                if (format=="nist") {
                        LLR=finalScore.getElement(0).toDouble();
                        outputResultNIST04Line(trainTypeTest,adaptationMode,segTypeTest,currentlinep->getElement(genderField),currentlinep->getElement(nameField),currentlinep->getElement(segField),currentlinep->getElement(decisionField),LLR,outFile);                              
                }
                
} // fin try

catch (Exception& e)
        { 
          cout << "FusionClasse :"<<e.toString().c_str() << endl;
        }
        return 0;
}

#endif //!defined(ALIZE_FusionScore_cpp)