Department of Electrical and Computer Engineering | Email: roman@jhu.edu |
The Johns Hopkins University | URL: http://bach.ece.jhu.edu/roman |
Barton Hall 105 | Tel (410) 243-2832 (H) |
3400 N. Charles St., Baltimore, MD 21218 | Fax (410) 516-5566 |
Analog and digital VLSI circuits, systems and algorithms for parallel signal processing and neural computation with application to pattern recognition, sound classification, autonomous systems design and low-power instrumentation.
Massively parallel mixed-signal VLSI processor implementing a Support Vector Machine (SVM) for high-dimensional data classification. Pattern recognition and acoustic classification systems utilizing SVM parallel processing chips. Low-power charge-mode parallel A/D converters, delta-sigma converters, and high-speed, high-resolution, digitally calibrated pipelined algorithmic A/D converters (with Northrop Grumman Corporation, Baltimore, MD). Reinforcement learning algorithms for intelligent robot navigation. Low-power analog and digital VLSI chips for high-performance instrumentation. Xilinx FPGA-based PCI bus interfaces for video and image processing microchips. Linux and Win2000 C++ device drivers, and Matlab interfaces.
Developed Q-learning-based algorithms for mobile micro-robot navigation. Investigated their mapping onto VLSI architectures.
Hardware-software co-design with Verilog, Xilinx tools, PALASM. Design and development of electrical and computer systems for new generation color copiers.
Design and simulation of digital cells for ASICs, and design of layouts for different ASIC implementations. Created a Perl extraction tool to automate the process of getting data from HSPICE output files into a condensed and useful state used for creating of Verilog ASIC models. Analysis time has been reduced by 10-15 times.