Initial Commit

AUTHORS

@@ -0,0 +1,4 @@
+Fabio Valente ([email protected])
+Deepu Vijayasenan ([email protected])
+David Imseng ([email protected])
+Srikanth Madikeri ([email protected])

INSTALL

@@ -0,0 +1 @@
+For installation instructions, see src/diarization/README

README

@@ -0,0 +1,39 @@
+This README file explains, in brief, the commands to 
+compile and run the IB diarization toolkit.
+
+# PREREQUISITES
+
+There are three prerequisites to the package
+
+1) Cmake -- required to build the package
+2) libboost -- boost library for C++
+3) openmp library -- for multithreaded processing
+
+# CLEAN COMPILATION
+
+$> cd src/diarization/cmake/
+$> cmake .
+$> # make clean if cmake has already been run once
+$> make
+
+# TO RUN
+
+$> # cd to $IB_DIARIZATION_HOME
+$> bash scripts/run.diarizeme.sh ipfile scpfile opfolder fileid [betaval]
+
+
+Example command
+
+$> bash scripts/run.diarizeme.sh data/mfcc/AMI_20050204-1206.fea data/scp/AMI_20050204-1206.scp result.dir/ AMI_20050204-1206
+
+To test the result use md-perl-eval tool available on the NIST website
+
+$> perl md-eval-v21.pl -m -afc -c 0.25 -r data/rttm/AMI_20050204-1206.rttm -s result.dir/AMI_20050204-1206.rttm
+
+The expected DER is 8.79%
+
+To add TDOA features, use the following script
+
+$>bash scripts/run.diarizeme.tdoa.sh data/mfcc/$AMI_20050204-1206.fea 0.8 data/tdoa/$AMI_20050204-1206.fea 0.2 data/scp/$AMI_20050204-1206.scp result.dir.tdoa $AMI_20050204-1206
+
+The expected DER is 7.12%

scripts/run.diarizeme.qsub.sh

@@ -0,0 +1,35 @@
+diarizeme_bin=/idiap/temp/msrikanth/repo/idiap-aib-diarization-dimseng/src/diarization/cmake/diarizeme
+beta_val=10
+
+numargs=3
+if [ $# -lt $numargs ]; then
+    echo "Usage: bash run.diarizeme.sh iplist scpfolder opfolder"
+    exit
+fi
+
+iplist=$1
+scpfolder=$2
+opdir=$3
+
+export nthreads=1
+
+if [ -e $opdir -a ! -d $opdir ]; then
+    echo "Warning: $opdir exists but is not a directory"
+    exit
+elif [ ! -d $opdir ]; then
+    mkdir $opdir
+fi    
+
+for fname in `cat $iplist`; do
+    fileid=`basename $fname .fea`
+    scpfile=$scpfolder/$fileid.scp
+echo "$diarizeme_bin \
+    --mfcc $fname 1.0 \
+    --recid $fileid \
+    --outdir $opdir \
+    --tmpdir $opdir \
+    -s $scpfile \
+    --beta $beta_val \
+    --nthread 1 > $opdir/$fileid.out" > temp.$fileid.sh
+qsub -l q1d temp.$fileid.sh     
+done

scripts/run.diarizeme.sh

@@ -0,0 +1,50 @@
+trap "exit" SIGINT
+
+
+diarizeme_bin=src/diarization/cmake/diarizeme
+beta_val=10
+
+numargs=4
+if [ $# -lt $numargs ]; then
+    echo "Usage: bash run.diarizeme.sh ipfile scpfile opfolder fileid [betaval]"
+    echo ""
+    echo "ipfile :  Input file with MFCC features in HTK format"
+    echo "scpfile : file with speech boundaries in the following format"
+    echo "          segment_name=file_name[start_frame,end_frame]"
+    echo "outdir :  directory to store temporary files and output rttm in"
+    echo "fileid :  ID of the file (to be used in rttm output)"
+    echo "betaval   :  (optional) beta value for IB clustering"
+    exit
+fi
+
+BASEDIR="$( cd "$(dirname "$0")" ; pwd -P )"
+export LD_LIBRARY_PATH=$BASEDIR/../lib/
+
+fname=$1
+scpfile=$2
+opdir=$3
+fileid=$4
+
+if [ $# -eq 5 ]; then
+    beta_val=$5
+fi
+
+if [ -z $nthreads ]; then
+    export nthreads=3
+fi
+
+if [ -e $opdir -a ! -d $opdir ]; then
+    echo "Warning: $opdir exists but is not a directory"
+    exit
+elif [ ! -d $opdir ]; then
+    mkdir -p $opdir
+fi    
+
+$diarizeme_bin \
+    --mfcc $fname 1.0 \
+    --recid $fileid \
+    --outdir $opdir \
+    --tmpdir $opdir \
+    -s $scpfile \
+    --beta $beta_val \
+    --nthread $nthreads > $opdir/$fileid.out

scripts/run.diarizeme.sh~

@@ -0,0 +1,54 @@
+trap "exit" SIGINT
+
+
+diarizeme_bin=src/diarization/cmake/diarizeme
+beta_val=10
+
+numargs=4
+if [ $# -lt $numargs ]; then
+    echo "Usage: bash run.diarizeme.sh ipfile scpfile opfolder fileid [betaval]"
+    echo ""
+    echo "ipfile :  Input file with MFCC features in HTK format"
+    echo "scpfile : file with speech boundaries in the following format"
+    echo "          segment_name=file_name[start_frame,end_frame]"
+    echo "outdir :  directory to store temporary files and output rttm in"
+    echo "fileid :  ID of the file (to be used in rttm output)"
+    echo "betaval   :  (optional) beta value for IB clustering"
+    exit
+fi
+
+BASEDIR="$( cd "$(dirname "$0")" ; pwd -P )"
+#export LD_LIBRARY_PATH=$BASEDIR/../lib/
+LD_LIBRARY_PATH=$BASEDIR/../lib/
+
+echo $LD_LIBRARY_PATH
+ls $LD_LIBRARY_PATH
+
+fname=$1
+scpfile=$2
+opdir=$3
+fileid=$4
+
+if [ $# -eq 5 ]; then
+    beta_val=$5
+fi
+
+if [ -z $nthreads ]; then
+    export nthreads=3
+fi
+
+if [ -e $opdir -a ! -d $opdir ]; then
+    echo "Warning: $opdir exists but is not a directory"
+    exit
+elif [ ! -d $opdir ]; then
+    mkdir -p $opdir
+fi    
+
+$diarizeme_bin \
+    --mfcc $fname 1.0 \
+    --recid $fileid \
+    --outdir $opdir \
+    --tmpdir $opdir \
+    -s $scpfile \
+    --beta $beta_val \
+    --nthread $nthreads > $opdir/$fileid.out

scripts/run.diarizeme.tdoa.sh

@@ -0,0 +1,63 @@
+trap "exit" SIGINT
+
+diarizeme_bin=src/diarization/cmake/diarizeme
+beta_val=10
+
+numargs=7
+if [ $# -lt $numargs ]; then
+    echo -ne "Usage: bash run.diarizeme.sh mfccfile mfccwt"
+    echo  " tdoafile tdoawt scpfile opfolder fileid [betaval]"
+    echo ""
+    echo "mfccfile :  Input file with MFCC features in HTK format"
+    echo "mfccwt   :  a value between 0.0 and 1.0 as a weight on MFCC features"
+    echo "tdoafile : Input file with TDOA features in TDOA format"
+    echo "tdoawt   : a value between 0.0 and 1.0 as a weight on TDOA features"
+    echo "scpfile : file with speech boundaries in the following format"
+    echo "          segment_name=file_name[start_frame,end_frame]"
+    echo "outdir :  directory to store temporary files and output rttm in"
+    echo "fileid :  ID of the file (to be used in rttm output)"
+    echo "betaval   :  (optional) beta value for IB clustering"
+    exit
+fi
+
+fname=$1
+fwt=$2
+tdoafname=$3
+tdoawt=$4
+scpfile=$5
+opdir=$6
+fileid=$7
+
+if [ $# -eq 8 ]; then
+    beta_val=$8
+fi
+
+export nthreads=1
+
+if [ -e $opdir -a ! -d $opdir ]; then
+    echo "Warning: $opdir exists but is not a directory"
+    exit
+elif [ ! -d $opdir ]; then
+    mkdir $opdir
+fi    
+
+echo $diarizeme_bin \
+    --mfcc $fname $fwt \
+    --tdoa $tdoafname $tdoawt \
+    --recid $fname \
+    --outdir $opdir \
+    --tmpdir $opdir \
+    -s $scpfile \
+    --beta $beta_val \
+    --nthread 1 
+
+
+$diarizeme_bin \
+    --mfcc $fname $fwt \
+    --tdoa $tdoafname $tdoawt \
+    --recid $fileid \
+    --outdir $opdir \
+    --tmpdir $opdir \
+    -s $scpfile \
+    --beta $beta_val \
+    --nthread 1  > $opdir/$fileid.out

src/aIB/include/aIB.h

@@ -0,0 +1,146 @@
+// Copyright (c) 2015 Idiap Research Institute, http://www.idiap.ch/
+// Written by Fabio Valente <[email protected]>
+// Written by Deepu Vijayasenan <[email protected]>
+// Written by David Imseng <[email protected]>
+// Written by Srikanth Madikeri <[email protected]>
+//
+// This file is part of the IB Speaker Diarization Toolkit.
+//
+// IB diarization toolkit is free software: you can redistribute it
+// and/or modify it under the terms of the GNU General Public License
+// version 3 as published by the Free Software Foundation.
+//
+// The IB Speaker Diarization Toolkit 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 General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with the IB Speaker Diarization Toolkit. If not, see
+// <http://www.gnu.org/licenses/>.
+
+
+#ifndef AIB_H 
+#define AIB_H
+
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <memory.h>
+#include <iostream>
+#include <vector>
+#include <map>
+#include "tools.h"
+
+using namespace std;
+using namespace libtools;
+
+
+
+struct bestmerge {
+  int bm_l;
+  int bm_r;
+};
+
+
+
+class Prm {
+ public:
+
+  Prm();
+
+  Prm(unsigned int xx,unsigned int yy,double Bbeta, int isUniformpriors);
+
+  Prm(unsigned int xx,unsigned int yy,double Bbeta, int isUniformpriors, vector <int>);
+
+  Prm(vector < vector <double> > *MM,double Bbeta, int isUniformpriors, vector <int> );
+
+  ~Prm();
+
+  unsigned int getX(); 
+  unsigned int getY();
+  double getBeta();
+  int getUniformPriors();
+  const vector <int> * getTsizes();
+
+ protected:
+  unsigned int X,Y;
+  double beta;
+  vector <int>* Tsizes;
+  int Uniformpriors;
+};
+
+
+class Inp {
+ public:
+  Inp(vector < vector <double> > * M, int Uniform);
+  ~Inp();
+  double getI();
+  double getHx();
+  double getHy();
+  const vector <vector <double> > *get_Py_x();
+  const vector <double> *get_Px();
+
+ protected:
+  vector < vector <double> > *Py_x;
+  vector < vector <double> > *Pxy;
+  vector <double> *Py;
+  vector <double> *Px;
+  MIvalues initialmi;
+};
+
+
+class TmpT{
+ public:
+  TmpT(Inp * input, Prm * currprm);
+  ~TmpT();
+  int get_tmpT_size();
+  vector<double>* get_Py_t(int ii);
+  vector<int>* get_Pt_x();
+  double get_Info_Ity(int ii);
+  double get_Pt(int ii);
+  int get_tmpt_size();
+  double get_Info_Ht(int ii);
+  void updatemergelog(unsigned int thismerge, int l,int r);
+  void updateDelta(unsigned int thismerge,double deltavalue);
+  void updateclustering(int thisr, int newval);
+  void updatePy_t(int thisl, int thisr);
+  void updatePt(int thisl, int thisr);
+  void updateInfoValues(int thisl, int thisr,double beta, double thisdelta);
+  void updateDeltaL(vector< vector<double> >* delta, TmpT* loctmpt, int thisl, int thisr, double beta,functionals ff);
+
+ protected:
+  vector <vector <int> > *MergeLog;
+  vector <double>  *Info_Ity;
+  vector <double>  *Info_Ht;
+  vector <double>  *Info_deltaL;
+  //AAACTHUNG tmpt_size is the index of the last element in the vectors/matrix not its size - this mess up is because in the original MATLAB notation vectors start from 1 - so let's stick to Slonim ...
+  int tmpt_size;
+  int prevTsize;
+  double pi_l,pi_r;
+  vector <int>  *Pt_x;
+  vector < vector <double> >  *Py_t;
+  vector <double>  *Pt;
+};
+
+
+
+class TT_elem {
+ public:
+  TT_elem(TmpT *r_tmpt, Inp *r_inp, Prm * r_prm);
+  double get_Ity();
+  vector <int> * get_Pt_x();
+  ~TT_elem();
+
+ protected:
+  vector <int>  *Pt_x;
+  double Ity;
+  double Ht;
+  double Ity_div_Ixy;
+  double Ht_div_Hx;
+  int size;
+};
+
+
+#endif