As a college student in Bucharest, Catalin Grigoras was fascinated by digital signals and what he could do with them. People around him said he was wasting his time, but he persevered. Now his work with evidence authentication is used around the world to help solve criminal cases. After earning a master's degree in electric engineering and a Ph.D. in digital signal processing from Politehnica University of Bucharest, he worked as a forensic expert with the Romanian Intelligence Service and the Ministry of Justice in Romania. In September 2010, he became director of the National Center for Media Forensics at the University of Colorado Denver.
The center offers the only master's level program in audio and video forensics in the country. Grigoras, his colleagues and students at the center use scientific methods to analyze audio and video evidence that comes into question during criminal trials. For instance, law enforcement agencies might want to determine whether a video or sound recording has been altered or edited in any way, or if a voice or face on a recording is authentic.
Grigoras was chairman of the European Network of Forensic Science Institutes – Forensic Speech and Audio Analysis Working Group from 2007-2009. He is a member of the Audio Engineering Society Subcommittee on Forensic Audio and the International Association of Forensic Phonetics and Acoustics. He has published numerous forensic audio/video articles and is a co-author of Best Practice Guidelines for ENF Analysis in Forensic Authentication of Digital Evidence.
Early on, "I did what I did because I didn't know it was impossible," Grigoras said. "Now the international recognition and validation of my work" is an honor.
Even away from work, Grigoras' inquisitive nature never wanes. He loves to travel, read, take photographs and discover new places. And he collects coins. "I like to study them. It's fascinating to see how much history can be embedded in one coin."
— Cynthia Pasquale
1. How did you choose this career?
I can say that the career chose me. It began as a game where I was trying to filter out different noises and measure with high accuracy a variety of signals. I was a student spending time in the lab and discovering the magic world of digital signals. Then it became like a hobby, where I was finding different phenomena and simultaneously learning more about forensic media science. People around me were asking me to stop wasting my time, but I continued to work, trying to push the scientific limits, knowing what I could do.
2. What are the functions of the National Center for Media Forensics and what are its goals?
The center offers a master's of science in recording arts with an emphasis in media forensics (MSRA-MF) degree program, and workshops for forensic experts from all over the world. We also do scientific research with our students, and prepare them for research, teaching or forensic practice in the investigative application of scientific methods aiding the criminal and civil justice systems.
The course topics that I cover with my colleague Jeff Smith are best practices in forensic media and litigation, forensic audio enhancement, forensic audio authentication, speaker recognition, forensic image enhancement, forensic image authentication and facial comparisons. We also teach MATLAB (a technical computing language) and provide our students with the know-how to develop forensic tools that currently are not available on the market. The most important scientific projects we are working on are the Electric Network Frequency (ENF) method for forensic authentication of digital audio/video recordings, and the digital image authentication procedure, where we plan to develop and propose a complex forensic methodology.
ENF represents a complex phenomenon that leaves nonrepetitive and nonpredictable traces on digital audio/video recordings. It's like a unique time fingerprint that remains in different digital evidence, allowing the forensic expert to verify or identify with high accuracy the date and time a recording was made, detect traces of manipulation or copy, and even to geolocate, in some cases, the place the recording was made. We can also call it forensic media archeology since the forensic expert is able to act like an archeologist who finds and dates artifacts.
3. How often do you collaborate with law-enforcement agencies and what kind of work do you do for them?
As forensic experts, we have to keep the confidentiality of our cases and clients. But generally, the scientific collaboration involves exchange of experience, implementation of new methods, development of best practice guidelines or standard operating procedures, and offering forensic expertise.
I've been asked to help with different cases, in different countries, involving complex digital media analysis, such as authentication and enhancement, speaker recognition and facial comparisons. I work with my colleague Jeff Smith and most of the cases we usually enter concern corruption or other criminal offenses.
4. What are some of the biggest advances that have been made in media forensics in the past decade or so?
The biggest developments have come about because of digital technology. Analog equipment and recordings are obsolete now, and everything around us is digital. Forensic media followed the evolution of modern society. We benefit from higher quality and longer recordings, can now build and search huge databases, and use automatic systems to process and analyze evidence. The digital revolution also raises new challenges for the forensic expert since most of the methods and theories developed for classic criminalistics are now obsolete.
5. One of your areas of research is forensic speaker recognition. How is it used, and what are your other personal areas of interest?
It's hard to separate the NCMF's and my personal projects because most of them coexist. Right now we are working to improve the ENF methodology, and to build and to propose high-resolution databases for forensic media labs. We also have a very challenging project under way that concerns image authentication. In the meantime, I'm also working on the Forensic Image Analysis System, in order to add and optimize new functions for digital image authentication. The system already is installed on the NCMF computers, and our students have been taught to use it and have experience in digital image authentication.
Automatic speaker recognition is a major domain of forensic audio and is the result of the past 20 to 30 years of digital evolution. By building voice sample databases, extracting voice models from evidence recordings and applying statistics, it is possible to automatically recognize speakers in a very short amount of time, which can be crucial in the fight against organized crime or terrorism.
Want to suggest a faculty or staff member for Five Questions? Please e-mail Jay.Dedrick@cu.edu