December 19, 2007
DRS Technologies, Inc. sponsored a Best Project Award competition for Digital Communication and Signal Processing Systems Design, taught by David Casasent, George Westinghouse Professor of ECE, and Associate Teaching Professor of ECE Tom Sullivan. DRS employees Jeff Silvey (M.S. 1998; B.S. 1996) and Patrick Mitrik from the company's Washington Operations traveled to campus to watch the final oral reports and demonstrations.
This fall, two groups shared the prize—their projects were "Content-Aware Image Resizing" and "The Sweet, Soothing Sounds of Pittsburgh." Formal awards will be presented by the company on Jan. 30 at a joint HKN/IEEE meeting on campus. According to its website, DRS Technologies is a leading supplier of integrated products, services and support to military forces, intelligence agencies, and prime contractors worldwide.
Digital Communication and Signal Processing Systems Design is a capstone design course. It provides a design and application hardware project experience in digital communications and signal processing systems using digital signal processing (DSP) hardware. Each of the course's groups completes a semester-long project of their choice and must implement it on a Texas Instruments (TI) C67 DSP-based development platform called a DSP Starter Kit (DSK).
ECE students Manuel Gonzalez-Rivero, Jim He, and Ziv Wolkowicki worked on a content-aware image resizing project, which implemented an algorithm that preserved important content in images while resizing. Standard methods for resizing images tend to warp and distort parts of the image, such as faces and text, if the aspect ratio of the original image is not preserved. However, the classmates used a relatively new algorithm called "seam carving,"  where a "seam" or path of pixels through the image of relatively low energy (of fairly constant image content) is removed at each iteration of the algorithm. Blindly removing full rows or columns of pixels from the image often removes important content; this new algorithm works its way around important content regions of the image to preserve its features.
Casasent noted that although resizing faces and text can be especially challenging, "the group found clever ways of preserving these features in their algorithm by automatically identifying important areas and protecting them from distortion by the seam carver...They implemented all major components of the project on the TI DSK and achieved very convincing results on some difficult images to resize."
Group members Umpei Kurokawa, Ryan Sherwin, Fabian Weissenberger, and Tian Yang visited Pittsburgh's Carnegie Music Hall, Carnegie Mellon University's Kresge Auditorium, and one of their own living rooms to measure impulse responses for their project on the sounds of the city. The team applied the room impulse response of the various locations to an input audio stream, implementing a zero-latency, real-time convolution reverberation algorithm.
Concert halls with a relatively long reverb time have correspondingly long room impulse responses. This leads to a noticeable latency (delay) in the output of a processed signal compared to the incoming audio if a basic time-domain convolution is used to process the audio stream. The zero-latency algorithm the team used was an implementation of an algorithm  in which the first set of samples of incoming audio were processed in real-time with a very short delay (that the human ear could not notice), and the remainder of the long impulse response of the room was broken up into segments and processed in the frequency domain.
"The algorithm requires attention to scheduling computations of different parts of the convolution in frequency, while servicing the immediate time-domain sections which gain priority through interrupts," said Casasent. "Their implementation had the TI DSK scheduled very efficiently to take advantage of almost all of the unit's processing power."
Digital Communication and Signal Processing Systems Design is one of the department's capstone design courses in the signals and systems area. Undergraduates in the ECE Department are required to take a capstone design course, which provides hands-on experience that enhances the student's repertoire of professional problem-solving and engineering design skills in the context of realistic engineering situations. Pupils work in teams to formulate a problem, propose an engineering solution or a design while weighing technical and socioeconomic constraints, and make sound professional judgments among alternative solutions.
 Avidan, S., and Shamir, A. (2007). "Seam Carving for Content-Aware Image Resizing," ACM Transactions on Graphics, Volume 26, Number 3, SIGGRAPH 2007.
 Gardner, W. G. (1994). "Efficient Convolution Without Input-Output Delay," Presented at the 97th Convention of the Audio Engineering Society, San Francisco. Preprint 3897.
ECE alumnus Jeff Silvey (right), a principal member of the technical staff at DRS Technologies, returned to campus to see the final projects for Digital Communication and Signal Processing Systems Design. ECE faculty member David Casasent is on the left.
Students working on a content-aware image resizing project implemented an algorithm that preserved important content in images while resizing. Resizing faces can be challenging, but the group found clever ways of preserving them in their algorithm.
“The Sweet, Soothing Sounds of Pittsburgh” project sought to replicate the acoustical characteristics of any concert hall or location across the globe through real-time convolution reverberation.