About

Roman Kuc is a Professor of Electrical Engineering at the School of Engineering & Applied Science at Yale University. He received his BSEE from the Illinois Institute of Technology in Chicago, IL, and his PhD in Electrical Engineering from Columbia University in New York, NY. His early career included work at Bell Laboratories, where he investigated efficient digital speech coding techniques. As a postdoctoral research associate at Columbia University, he applied digital signal processing to diagnostic ultrasound signals to characterize liver disease. At Yale, he serves as the Director of the Intelligent Sensors Laboratory, where he pursues research in biomimetic sensors for robotics and bioengineering. Professor Kuc has published over 200 papers in digital signal processing, robotics, and biomimetic sonar sensing, and is the author of four textbooks, including 'Electrical Engineering in Context' and 'The Digital Information Age.' He co-authored the Sonar chapter in the second edition of the Springer Handbook of Robotics with Lindsay Kleeman. His honors include an honorary Doctorate from the Hlushkov Cybernetics Institute of the Ukrainian Academy of Sciences, membership in the Connecticut Academy of Science and Engineering, fellowship in the Shevchenko Scientific Society, and recognition as an honorary Academician of the Ukrainian Academy of Sciences. He is also the first recipient of the Yale Sheffield Distinguished Teaching Award and has received an IEEE Acoustics, Speech and Signal Processing Society Paper Award.

Research topics

• Artificial Intelligence
• Computer Science
• Acoustics
• Physics
• Telecommunications
• Mathematics
• Speech recognition
• Computer vision

Selected publications

• My Encounters with Richard K. Chang

  WORLD SCIENTIFIC eBooks · 2025-06-22

  book-chapter1st authorCorresponding
  PublisherDOI

• Binaural echo error detection gives self-supervised learning to improve landmark classification

  The Journal of the Acoustical Society of America · 2024-10-01

  article1st authorCorresponding

  Binaural processing is typically associated with source localization using interaural time and level differences. This paper describes its role for improving landmark classification from an echo sequence. A brain-inspired system explores the blind human echolocation problem of differentiating two foliage targets with different sized leaves using audible echoes. A biomimetic sonar views each target by producing left and right-ear monaural echo waveforms whose target-specific frequency power spectra are classified using template matching. Binaural processing implements error detection by ignoring views that give contradictory monaural classifications. Binaural classification using consistent monaural classifications also produces self-labeled data that update monaural template estimates to continually improve target classification accuracy through additional target encounters. Monaural classification error probability starts at 0.225 after supervised learning with 10 encounters with each known target and reaches 0.134 after an additional 90 self-supervised template updates. The binaural classification error probability starts at 0.032 with an average of 1.49 binaural views required to achieve consistent monaural classifications and reaches 0.019 with an average of 1.32 views.

  PublisherDOI

• Brain-inspired sensorimotor echolocation system for confident landmark recognition

  The Journal of the Acoustical Society of America · 2022-09-01 · 1 citations

  article1st authorCorresponding

  A landmark is a familiar target in terms of the echoes that it can produce and is important for echolocation-based navigation by bats, robots, and blind humans. A brain-inspired system (BIS) achieves confident recognition, defined as classification to an arbitrarily small error probability (PE), by employing a voting process with an echo sequence. The BIS contains sensory neurons implemented with binary single-layer perceptrons trained to classify echo spectrograms with PE and generate excitatory and inhibitory votes in face neurons until a landmark-specific face neuron achieves recognition by reaching a confidence vote level (CVL). A discrete random step process models the vote count to show the recognition probability can achieve any desired accuracy by decreasing PE or increasing CVL. A hierarchical approach first classifies surface reflector and volume scatterer target categories and then uses that result to classify two subcategories that form four landmarks. The BIS models blind human echolocation to recognize four human-made and foliage landmarks by acquiring suitably sized and dense audible echo sequences. The sensorimotor BIS employs landmark-specific CVL values and a 2.7° view increment to acquire echo sequences that achieve zero-error recognition of each landmark independent of the initial view.

  PublisherDOI

• Brain-Based Classification of Landmarks using Audible Sonar Echo Sequences

  2021 11th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS) · 2021 · 3 citations

  1st authorCorresponding
  • Artificial Intelligence
  • Computer Science
  • Artificial Intelligence

  A brain-based system (BBS) is described that models blind human echolocation to classify four environmental landmarks. Two target groups include surface reflectors (SR) and volume scatterers (VS) and each group contains two target types. SR includes single-post targets modeling isolated reflectors, and multiple-post targets modeling multiple strong reflectors and reverberation. The VS group includes two plants having leaf areas that differ by a factor of four. Each target produced 5,600 echoes that were transformed into spectrograms to display the spectral variation over a 46 cm range interval. A database of 22,400 entries of target views in 0.9° increments provided data corresponding to probing a target from a specified set of sequential views. The BBS implements a brain predictive model with three binary single-layer regression perceptrons followed by accumulators that produce averages from sequential echoes for classification. The BBS generalization classification accuracy from single echoes exceeds 95% and improves with the number and the view increment of sequential echoes. The BBS system provides a sequential scanning strategy to acquire sufficient data to achieve errorfree classification.

  PublisherDOI

• Artificial neural network classification of foliage targets from spectrograms of sequential echoes using a biomimetic audible sonar

  The Journal of the Acoustical Society of America · 2020 · 6 citations

  1st authorCorresponding
  • Artificial Intelligence
  • Computer Science
  • Artificial Intelligence

  Classifying foliage targets using echolocation is important for recognizing landmarks by bats using ultrasonic emissions and blind human echolocators (BEs) using palatal clicks. Previous attempts to classify foliage used ultrasonic frequencies and single sensor (monaural) detection. Motivated by the echolocation capabilities of BEs, a biomimetic sonar emitting audible clicks acquired 5600 binaural echoes from five sequential emissions that probed two foliage targets at aspect angles separated by 18°. Echo spectrograms formed feature vector inputs to artificial neural networks (ANNs) for classifying two targets, Ficus benjamina and Schefflera arboricola, with leaf areas that differ by a factor of four. Classification performances of ANNs without and with hidden layers were analyzed using tenfold cross-validation. Performance improved with input feature size, with binaural echo classification outperforming that using monaural echoes for the same number of emissions and for the same number of echoes. Linear classification accuracy was comparable to that using nonlinear classification with both achieving fewer than 1% errors with binaural spectrogram features from five sequential emissions. This result was better by a factor of 20 compared to previous classification of these targets using only the time envelopes of the same echoes.

  PublisherOA PDFDOI

• Artificial neural network classification of surface reflectors and volume scatterers using sequential echoes acquired with a biomimetic audible sonar

  The Journal of the Acoustical Society of America · 2020 · 6 citations

  1st authorCorresponding
  • Artificial Intelligence
  • Computer Science
  • Artificial Intelligence

  = 0.220, indicating that envelope features by themselves are inadequate to accurately differentiate foliage targets.

  PublisherDOI

• Teaching The Non Science Major: E El0l The Most Popular Course At Yale

  2020-09-01 · 1 citations

  article1st authorCorresponding

  Abstract NOTE: The first page of text has been automatically extracted and included below in lieu of an abstract Session 2532 Teaching the non-science major: EEl0l - The most popular course at Yale Roman Kuc Department of Electrical Engineering Yale University, New Haven, CT 06520-8284 EE 101 - The Digital Information Age, a course for non-science majors, is the largest course at Yale with an enrollment of more than 500 students. The goal of the course is to describe how common-place information systems work and why they work that way by illustrating clever engineering solutions to technical problems. The course considers the following topics: information sources, logic gates, computer hardware and software, measuring information using entropy, information coding and encryption, information transmission and information manipulation. EElOl includes a hardware and software project. For the hardware project each student implements a bean counter that counts a student-specific number of beans. The real success of the course is the software project that involves writing a personal World Wide Web page and developing a Web page for a Yale-affiliated organization. Having taken the course, students feel that they have an appreciation for the digital information artifacts they encounter on a daily basis. The joys and tribulations of teaching EElOl are discussed. Introduction The problems with teaching science and technology to the non-science major are well known [l, 2, 3,4]. The main problem is dealing with the wide spectrum of the student’s experience in math and the sciences. A secondary problem is what to have the students do that is meaningful, instructive and satisfying. The solution to both problems that has found acceptance at Yale is to present material that the students find interesting and relevant, thus providing the motivation for expending the effort to learn the material. EElOl is a course for non-science majors, as well as for freshmen are considering EE as a major. In addition to teaching students about electrical engineering, the student is invited to be an engineer for one semester: To think quantitatively, to design a simple digital system that does something useful and to develop pages on the World Wide Web. The course attempts to teach technology in the least stressful manner to allow the poets, who would not normally have access to this material, to take the course. Among the difficul- ties with teaching a 100-level course are that there are no prerequisites and the course itself is not usually a prerequisite for follow-on courses. In a student’s time and effort, it competes with courses in the major. Teaching such a course introduces a challenge to make the course accessible to the liberal arts major, while still making it interesting for the science major.

  PublisherDOI

• The evolution of bat robots

  The Journal of the Acoustical Society of America · 2019-03-01 · 1 citations

  article

  The extraordinary skills of bats in supporting dexterous mobility in complex environments based on just two pulsed trains of one-dimensional biosonar echoes has attracted attention from engineers for several decades already. To explore whether it is possible to reproduce at least certain of these capabilities, a diverse set of “bat robot” prototypes have been built. The earliest, most basic of these systems were limited to estimating the distance of sonar targets based on the acoustic time-of-flight. From there, systems improved to take advantage of more echo waveform features, e.g., for target recognition. Two (or more) receivers were introduced to exploit binaural differences, e.g., for target tracking. Rigid ear rotations served to enhance the signal-to-noise ratio be focusing on a target or to determine target direction from the echo amplitudes received across a scan. Biomimetic emission and reception baffle shapes, i.e., “noseleaves” and “pinna,” were added to narrow the sonar beams and create direction-depended spectral signatures. Deformations of flexible baffle structures that mimic the muscular actuation of the noseleaf and pinna shapes seen in bats have been added to these systems. Mobility of the entire systems has been provided by mounting them on pan-tilt units, robot arms, mobile robots, and drones.

  PublisherDOI

• Biomimetic sonar echo parameters form cognitive maps

  The Journal of the Acoustical Society of America · 2019-03-01

  article1st authorCorresponding

  A biomimetic audible sonar probes 2.5D targets and processes binaural echoes to extract values of eight parameters to generate two-dimensional cognitive maps. Targets are configured using posts connected by tangential planes. Being tuned to recognize posts and planes, the sonar produces a cognitive map that is composed of these two components. A platform with translational and rotational degrees of freedom employs right-ear dominance to implement a landmark-centric scanning trajectory whose step size adaptively changes with echo information. The sonar tracks the target by maintaining a constant first echo arrival time and equalizes binaural echo times to form singular echoes. When observed, singular echoes identify landmarks defined by post radii and locations. The mapping process employs five states from detection to termination that passes through the singular echo state. Separate states detect post pairs that exhibit echo interference and planes that exhibit echo amplitude differences. The scanning process terminates when the current landmark parameters match those of the first landmark. Two targets configured with three posts and an added plane illustrate the procedure.

  PublisherDOI

• Generating cognitive maps using echo features from a biomimetic audible sonar

  The Journal of the Acoustical Society of America · 2019-04-01 · 6 citations

  article1st authorCorresponding

  A sonar cognitive map displays target components that are specified by signal features extracted from a single binaural echo pair. A biomimetic audible sonar probes targets configured using posts connected by tangential planes. Echo envelopes are processed to extract values of eight parameters that govern the mapping process. Being tuned to recognize posts and planes, a cognitive map is composed of these two components using the posts' centers and radii as landmarks. A platform with translational and rotational degrees of freedom implements a landmark-centric scanning trajectory whose step size adaptively changes with echo information. The sonar tracks the target surface by maintaining a constant first-echo arrival time and by equalizing binaural echo times to form singular echoes that identify landmarks. The mapping process employs five states from detection to termination that pass through the singular echo state. Separate states process echo interference caused by two posts and echoes from planar surfaces. Sonar scanning stops when the current landmark parameters match those of the first landmark. Two targets configured with three posts and an added plane illustrate the procedure. Cognitive maps exhibit landmark locations that are accurate to ±5% with post radius estimates accurate to ±20%.

  PublisherDOI

Frequent coauthors

• Billur Barshan

  25 shared

• O. Bozma

  Yale University

  22 shared

• Rolf Müller

  14 shared

• M. Schwartz

  University of California, Los Angeles

  7 shared

• Lindsay Kleeman

  Monash University

  6 shared

• Victor Kuc

  Iona College

  4 shared

• V. Brian Viard

  Cheung Kong Graduate School of Business

  4 shared

• Gennadiy M. Bakan

  4 shared

Awards & honors

• Honorary Doctorate, Glushkov Institute of Cybernetics, Kyiv,…
• Award for Excellence in Physical Sciences and Mathematics, f…
• Elected to Connecticut Academy of Science and Engineering (2…
• Fellow, Shevchenko Scientific Society (2001)
• Order of the Golden Bulldog Award (1999)