00001 /* 00002 This file is part of LIA_RAL which is a set of software based on ALIZE 00003 toolkit for speaker recognition. ALIZE toolkit is required to use LIA_RAL. 00004 00005 LIA_RAL project is a development project was initiated by the computer 00006 science laboratory of Avignon / France (Laboratoire Informatique d'Avignon - 00007 LIA) [http://lia.univ-avignon.fr <http://lia.univ-avignon.fr/>]. Then it 00008 was supported by two national projects of the French Research Ministry: 00009 - TECHNOLANGUE program [http://www.technolangue.net] 00010 - MISTRAL program [http://mistral.univ-avignon.fr] 00011 00012 LIA_RAL is free software: you can redistribute it and/or modify 00013 it under the terms of the GNU Lesser General Public License as 00014 published by the Free Software Foundation, either version 3 of 00015 the License, or any later version. 00016 00017 LIA_RAL is distributed in the hope that it will be useful, 00018 but WITHOUT ANY WARRANTY; without even the implied warranty of 00019 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 00020 GNU Lesser General Public License for more details. 00021 00022 You should have received a copy of the GNU Lesser General Public 00023 License along with LIA_RAL. 00024 If not, see [http://www.gnu.org/licenses/]. 00025 00026 The LIA team as well as the LIA_RAL project team wants to highlight the 00027 limits of voice authentication in a forensic context. 00028 The "Person Authentification by Voice: A Need of Caution" paper 00029 proposes a good overview of this point (cf. "Person 00030 Authentification by Voice: A Need of Caution", Bonastre J.F., 00031 Bimbot F., Boe L.J., Campbell J.P., Douglas D.A., Magrin- 00032 chagnolleau I., Eurospeech 2003, Genova]. 00033 The conclusion of the paper of the paper is proposed bellow: 00034 [Currently, it is not possible to completely determine whether the 00035 similarity between two recordings is due to the speaker or to other 00036 factors, especially when: (a) the speaker does not cooperate, (b) there 00037 is no control over recording equipment, (c) recording conditions are not 00038 known, (d) one does not know whether the voice was disguised and, to a 00039 lesser extent, (e) the linguistic content of the message is not 00040 controlled. Caution and judgment must be exercised when applying speaker 00041 recognition techniques, whether human or automatic, to account for these 00042 uncontrolled factors. Under more constrained or calibrated situations, 00043 or as an aid for investigative purposes, judicious application of these 00044 techniques may be suitable, provided they are not considered as infallible. 00045 At the present time, there is no scientific process that enables one to 00046 uniquely characterize a persones voice or to identify with absolute 00047 certainty an individual from his or her voice.] 00048 00049 Copyright (C) 2004-2010 00050 Laboratoire d'informatique d'Avignon [http://lia.univ-avignon.fr] 00051 LIA_RAL admin [alize@univ-avignon.fr] 00052 Jean-Francois Bonastre [jean-francois.bonastre@univ-avignon.fr] 00053 */ 00054 00055 #if !defined(ALIZE_ComputeTest_h) 00056 #define ALIZE_ComputeTest_h 00057 00058 #include "alize.h" 00059 00060 int ComputeTest(alize::Config&); 00061 int ComputeTestFA(alize::Config&); /* DM version */ 00062 int ComputeTestJFA(alize::Config&) /* AL version */; 00063 int ComputeTestLFA(alize::Config&); /* AL version */ 00064 int ComputeTestDotProduct(alize::Config&); /* AL version */ 00065 int ComputeTestNAP(alize::Config&); 00066 int ComputeTestByLabel(alize::Config&); 00067 int ComputeTestHisto(alize::Config&); 00068 #endif //
1.7.2