Member Highlight: Dr. Sanjit K. Mitra, Distinguished Professor Emeritus of Electrical & Computer Engineering

Dr. Sanjit Kumar Mitra is an Distinguished Professor Emeritus of Electrical and Computer Engineering, University of California, Santa Barbara, and Stephen and Etta Varra Professor Emeritus of Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles. He is an Indian-American scientist and earned his PhD in electrical engineering at the University of California, Berkeley.  He has held visiting appointments at Austria, Australia, Brazil, Croatia, Germany, Japan, Finland, India, Turkey and United Kingdom. He has initially worked in analog signal processing. In the mid 1960s, he and the first to design an active RC lowpass filter containing operational amplifiers to replace the analog lowpass LC filters for sampling analog signals for use in the conversion from frequency-division multiplex communication systems to time-division multiple communication systems. In the late 1960s, he worked in the then emerging area of digital signal processing and was honored as a pioneer in signal processing at the 1998 IEEE International Conference on Acoustics, Speech, and  Signal Processing and honored as a pioneer in circuits, systems and signal processing at the 2017 IEEE International Symposium on Circuits and Systems.  He has served the IEEE in various capacities including service as the President of the IEEE Circuits and Systems Society. He is the recipient of the Society Award of the IEEE Circuits and Systems Society, the Technical Achievement Award and the Society Award of the IEEE Signal Processing Society and the Technical Achievement Award of the EURASIP (European Association for Signal Processing). Mitra has also been honored with the IEEE Gustav Kirchhoff Award and the IEEE James H. Mulligan Jr. Award. He is an Academician of the Academy of Finland, a Life Fellow of the IEEE, a member of the US National Academy of Engineering and 9 foreign academies. View Dr. Mitra’s full biography.

We approached Dr. Sanjit Mitra, with the following questions, to learn more.

1. Why did you choose to become faculty in the field of Signal Processing?

I came to UC Berkeley to work for my PhD degree in computer engineering as I had worked for two years as a maintenance engineer for India’s first computer at an institute in Calcutta. During this period, I learned machine language programming and wrote several computer games that could be played against the computer. I decided to go the United States for doctoral research in computer engineering at UC Berkeley. At that time, the professor who taught computer engineering courses was on a sabbatical leave. The electrical engineering department required that all PhD students must pass an 8-hour long screening examination consisting of questions based on materials of upper division courses in electrical engineering before they can start their doctoral research. As I was unsure of passing this examination, I decided to get a second Master’s degree. The department used to publish a short booklet containing a brief description of the project one of which was to design practical analog filters containing lossy passive elements using a computer proposed by a professor. Since I knew machine language programming, I decided to work on this project which led to my Master’s thesis and also resulted in a paper published in an IEEE Transactions. I continued working in the area of analog signal processing for doctoral research and developed several methods of designing inductor-less analog filters.

2. How does your work affect society? 

In the late 1960s, I switched my research interest to the emerging field of digital signal processing. The field matured and flourished in the 1960s and 1970s, and digital signal processing became widely used with specialized digital signal processor chips in the 1980s. The impact of signal processing on the society can be seen from the article published in the IEEE Signal Processing Magazine.

3. What challenges have you had to face to get to where you are today?

During my active teaching career, I have worked on all various problems in digital signal processing. One of the earliest problems being pursued by some researchers was the development of digital filter structures with low coefficient sensitivity on their frequency responses. Published works were based on the analysis of biquadratic structures. With one of my doctoral students, I developed the conditions for realizing low-sensitivity infinite impulse response (IIR) and finite impulse response (FIR) digital filter structures along with several realization methods.

One challenge I faced in the beginning of my teaching career was to convince my professional colleagues that it is important to develop algorithms with the least number of digital multipliers.  Even though signal processor chips were commercially available, they consume power and occupy a large amount of space on a practical fully integrated circuit. With my student and a visiting researcher, we developed digital filter structures with the least number of digital multipliers. Our work on the design of finite-impulse response (FIR) and infinite impulse response (IIR) digital filters with the fewest number of multipliers have been incorporated in two commercially available software packages: Signal Processing Toolbox of MATLAB (https://en.wikipedia.org/wiki/MATLAB) and LabVIEW (https://en.wikipedia.org/wiki/LabVIEW) of National Instruments.

Another challenge I faced was to convince some of my professional colleagues to think “outside the box” and go beyond the widely known and used Discrete Fourier Transform (DFT) after the publication of the Fast Fourier Transform (FFT) algorithm which computes the frequency domain representation of a finite number of discrete tine-domain samples.  FFT basically computes all frequency domain samples whether they are required or not in a practical problem. Together with visiting researchers and a student, I developed three generalizations of the popular digital Fourier transform (DFT) with each establishing new research directions: Subband DFT and Warped DFT and Nonuniform DFT.

Subband DFT (https://squ.pure.elsevier.com/en/publications/subband-dft-part-i-definition-interpretation-and-extensions) is based on a decomposition of the time-domain samples into a set of smaller length subsequences approximately separated in the spectral domain. The frequency-separation property of the subsequences permits elimination of the subsequences with negligible energy contribution from the DFT calculation, thus resulting in a faster approximate DFT computation.

Warped DFT (https://ieeexplore.ieee.org/document/948436) is the evaluation of frequency samples at nonuniformly spaced points on the unit circle and has been applied to spectral analysis, design of tunable FIR filters, resolving closely spaced sinusoids, frequency estimation of noise-corrupted short-length sinusoid, robust speech recognition, perceptual noise reduction system, perceptual speech enhancement and signal analysis.

Nonuniform DFT (https://en.wikipedia.org/wiki/Non-uniform_discrete_Fourier_transform) is the frequency domain representation of a finite-length sequence located at unequally spaced points. It has been applied to discrete multitone transmission, antenna array design, magnetic resonance imaging, numerical solution of partial differential equations and spectral analysis.

4. What advice would you give to scientists/engineers in signal processing?

Many of the projects that my students and I worked on were suggested by the industry and involved the processing of images and videos. For example, my student and I along with a visitor from Italy patented a method for equalization of practical (non-ideal) transmission channels which considerably simplified the decoding of corrupted data sequences. This patent has been licensed by Alcatel/Telettra Corporation and has been incorporated in one of their equipment.

I believe that it is imperative for scientists/engineers to engage with industry and identify the problems related to signal processing that are important to them and work on these problems. It is also important to work on non-traditional problems.