Gettting to Know: Dr. Miloš Doroslovački

 

 

Dr. Doroslovacki conducts research in signal processing, communication signals and systems, discrete-time signal and system theory, and wavelets and their applications. His recent projects have centered on adaptive wavelet-based  echo cancellation for voice transmission over digital networks, attitude rate estimation using GPS, techniques for automatic recognition of modulations, sparse system identification, distributed estimation, Bayesian learning, machine learning, and signal processing for hardware security.

 

Please describe in simple terms the research that you conduct.

My research is about mathematical algorithms that enable quality reception of speech, images, video, efficient transmission over communication channels, intelligent receivers, automatic recognition of patterns in radio communication signals, images, and hardware behavior.

 

What is the significance of your research in terms of practical applications?

They can reduce echo in voice communications over communication networks; they can use collaborative distributive processing to improve target detection; radio receivers can automatically tune to radio stations; they enable recognition of abnormal tissue states; they improve wireless simultaneous transmission of information and energy and can enable protection against stealing data in computer systems.

 

Which of your research areas is most interesting or most challenging to you, and why?

Challenges change with time, the technical interest moves fast in engineering fields.  Something that was impossible five years ago is possible today.  Currently, what is most challenging for me is understanding learning techniques for neural networks, improving their robustness and efficiency, and designing signals for efficient wireless simultaneous transmissions of information and energy.

 

How did you become interested in your field of research?

I was always attracted by mathematical thinking and the application of mathematical modeling in physics and engineering.  The signal processing field allows me to fulfill this interest of mine.

 

What are the challenges or limitations associated with this field of research?

There is no solution in the form you originally looked at.  Many problems are so complex that it is not possible to build an appropriate model for them.  One has to resort to reasonable approximations and to be led by intuition, and therefore one has to think out of the box to find a solution.

 

What are hoping to achieve or learn through your research?

Understanding of technical problems and being able to generate hypotheses that should be then checked analytically and/or experimentally.  To design techniques that will allow in an optimal way to get rid of signal distortions, to automatically recognize objects, such as in images, to improve energy efficiency.  In the end, practical applications should be improved or enabled.

 

Is there anything about your research that you think makes it unique from other research being conducted in this area?

This is probably a combination of understanding fundamentals of problems, doing as much as possible a strict mathematical analysis, and incorporating adaptability with respect to changing parameters of the systems considered.

 

What successes or milestones have you reached thus far in your research?

We produced a benchmark algorithm for echo cancellation for sparse echo paths, and also a fundamental limit for simultaneous localization in time and frequency for digital filters is found and used by practitioners of digital filter design.  Efficient, automatic modulation recognition techniques based on machine learning techniques, as well as improved training and interpretability of deep neural networks are achieved.  We developed unique and efficient techniques for detection and defense of hardware side channels.

 

Are you collaborating with anyone else?  Who funds your research?

I have recently collaborated with fellow ECE professor Dr. Guru Venkataramani on hardware security problems and with BME professor Dr.  Murray Loew on deep learning neural networks.  NSF, NSA, and industry funding—such as Texas Instruments—help some of the research I have mentioned.

 

What role do GW students play in your research and what do you gain from their involvement?

Research would not be possible without my PhD students.  Their energy and interest in contemporary technical issues motivates and guides me.  They learn from me, but I learn from them about things they see much better than I do.