Projects / Programmes
Forensic Speaker Identification
| Code |
Science |
Field |
Subfield |
| 2.07.07 |
Engineering sciences and technologies |
Computer science and informatics |
Intelligent systems - software |
| Code |
Science |
Field |
| T121 |
Technological sciences |
Signal processing |
| P170 |
Natural sciences and mathematics |
Computer science, numerical analysis, systems, control |
| P176 |
Natural sciences and mathematics |
Artificial intelligence |
Forensic speaker identification, forensic voice identification, speech technologies, natural language processing
Researchers (9)
| no. |
Code |
Name and surname |
Research area |
Role |
Period |
No. of publicationsNo. of publications |
| 1. |
20242 |
PhD Andraž Bežek |
Computer science and informatics |
Researcher |
2007 - 2009 |
25 |
| 2. |
15745 |
Robert Blatnik |
|
Technical associate |
2007 - 2009 |
75 |
| 3. |
05026 |
PhD Bogdan Filipič |
Computer science and informatics |
Researcher |
2007 - 2009 |
482 |
| 4. |
08501 |
PhD Matjaž Gams |
Computer science and informatics |
Researcher |
2007 - 2009 |
1,690 |
| 5. |
02571 |
Mitja Lasič |
Computer science and informatics |
Technical associate |
2007 - 2009 |
24 |
| 6. |
23581 |
PhD Mitja Luštrek |
Computer science and informatics |
Researcher |
2007 - 2009 |
476 |
| 7. |
23318 |
PhD Domen Marinčič |
Computer science and informatics |
Researcher |
2007 - 2009 |
31 |
| 8. |
15656 |
PhD Tomaž Šef |
Computer science and informatics |
Head |
2007 - 2009 |
394 |
| 9. |
24894 |
PhD Tea Tušar |
Computer science and informatics |
Researcher |
2007 - 2009 |
209 |
Organisations (1)
| no. |
Code |
Research organisation |
City |
Registration number |
No. of publicationsNo. of publications |
| 1. |
0106 |
Jožef Stefan Institute |
Ljubljana |
5051606000 |
90,753 |
Abstract
The modern knowledge-based society is founded on new information-communication technologies, which bring radical changes into the organization and activities of institutions. However, these achievements can also be used for illegal purposes. Well organized criminal associations have always been one step ahead of the public prosecutors. To uncover these organizations, the police make use of various special methods and instruments, like tapping and recording telephone calls. Later, in trial hearing, these pieces of evidence are frequently doubted of. Moreover, it is very difficult to prove that the recorded voice belongs to the defendant and not some other person. A great help in this process comes from the state-of-the-art methods for forensic speaker recognition.
The basic goal of this project is to gather a specialized database of pronunciations from telephone records (the speaking corpus of telephone conversations), which would be big enough and adequate to enable systematic research in the field of speaker identification. Our research aims at finding the speaking parameters, which are different for individual speakers and therefore serve for speaker identification. The long-term goal of this research will be to develop a system for automatic recognition of speakers for forensic purposes. The intention of the project is to increase the accessibility of pretentious and expensive investigations in the field of forensic speaker identification and to increase the reliability of the results. Special attention will be given to the specificity of Slovenian language.
With this project, the state will give the opportunity to contemporary scientific and technological means to be used more frequently for uncovering the organized crime in Slovenia. As a result, the productiveness will increase, as well as the efficiency of investigating agencies and Slovene courts. In this way, the state will accelerate some long-lasting and complicated legal matters. With a more effective persecution of organized crime, the safety of citizens and the quality of life will be increased. Furthermore, the results of research and development will be transmitted from theory into practice, where they will be applied to solve complicated and demanding social problems.
Significance for science
Applied science of the speaker recognition is becoming increasingly popular in many commercial products and services. Methods of speaker recognition must be adopted to various degradations imposed by transmission over telephone networks. Perceived quality of the voice transmitted over the telephone networks is traditionally being one of the most important indicators for the quality of the plain old telephone service. However, the introduction of packet based transmission in fixed and mobile telephony brings new issues, such are new coding mechanisms, packet loss concealment techniques, voice quality optimization and others that requires new voice quality measurement techniques. Therefore voice quality measurements are of great importance for the applied science of the speaker recognition and for the development of new reliable automated forensic speaker recognition systems.
The specialized database of pronunciations from telephone records (the speaking corpus of telephone conversations) enable systematic research in the field of speaker identification. Our research aims at finding the speaking parameters, which are different for individual speakers and therefore serve for speaker identification.
The long-term goal of the research is to develop a system for automatic recognition of speakers for forensic purposes. The development is mainly focused on research of linguistically dependant parts of the system. Namely, each language is specific and uncomparable to other languages through speech parameters. For linguistically independent parts, state-of-the-art findings are utilised.
Significance for the country
Speaker recognition research is fundamental for the development of the automatic forensic speaker recognition systems. The American National Institute of Standards and Technology each year organizes the NIST speaker recognition evaluation series to contribute to the direction of research efforts and the calibration of technical capabilities of text independent speaker recognition. The overarching objective of the evaluations has always been to drive the technology forward, to measure the state-of-the-art, and to find the most promising algorithmic approaches. Therefore it is essential that Slovenia follows American governments positive example by funding speaker recognition research projects with particular emphasis on the native language specifics, local law and practice of crime investigation in the future. Additionaly, forensic speaker recognition technologies have to adopt to new conditions enforced by many new voice communication technologies, e.g. influence of packet loss and jitter on the voice quality in packet switched networks. While objective voice quality measurements are continuously performed by the telephony service providers, the employment of such approaches in the research of speaker recognition robustness is bringing new knowledge that is complementary applicable not just in maintenance and development of the telecommunication infrastructure, but also more importantly in the development of robust, more efficient and trustworthy forensic speaker recognition systems in the future.
With the support of this research work the state gives the opportunity to contemporary scientific and technological means to be used more frequently for uncovering the organized crime in Slovenia. As a result, the productiveness increase, as well as the efficiency of investigating agencies and Slovene courts. With a more effective persecution of organized crime, the safety of citizens and the quality of life is increased. Furthermore, the results of research and development are transmitted from theory into practice, where they are applied to solve complicated and demanding social problems.
Most important scientific results
Annual report
2008,
final report,
complete report on dLib.si
Most important socioeconomically and culturally relevant results
Annual report
2008,
final report,
complete report on dLib.si
