 | TI announces winner |
On May 7, 1996, Texas Instruments announced that Dilip Krishnan
and Showbhik Kalra, both students at Singapore's Nanyang Technological
University (NTU), will share the US $100,000 grand prize in TI's
DSP Solutions Challenge, a contest that encourages engineering
students to use DSPs for new applications.
The Challenge, the first ever worldwide contest of its kind, received entries from more than 230 teams representing more than 700 students in 26 countries. The three finalist teams were from NTU, the University of Maryland in the United States, and Ecole Francaise d'Electronique et d'Informatique (EFREI) of France.
"Every country, without exception, produced visionary projects," said Torrence Robinson, TI's North American university program manager. "The success of the Challenge shows that, equipped with raw technology, students can design amazing new applications and learn DSP fundamentals in the process."
Challenge teams were judged on overall design creativity and applicability to real world needs that affect the way people live, learn, work, and play.
Challenge winners Dilip Krishnan and Showbhik Kalra are second-year computer engineering students at Nanyang Techno-logical University (NTU), in the southwest of Singapore. Using parallel process- ing techniques on a network of TMS320C40 DSPs, the pair developed a motion-picture restoration system which erases damaged areas of old movie film, automatically filling in the resulting empty space with information captured from elsewhere in the film. Because movie image frames do not change significantly from one frame to the next, frames preceding and succeed- ing a damaged image provide enough comparative information to enable the system to detect degraded areas, mathema-tically model the image region, and fill in the degraded region with a restored image.
The restoration of a classic motion picture is a labor-intensive, costly undertaking. For example, the restoration of Snow White took 18 weeks with 60 workstation operators using 40 workstations in three shifts a day, seven days a week. NTU's DSP-based system will allow more advanced techniques that could handle such projects in a fraction of the time, at much less cost, and with far fewer people.
Both students said that this project will not end with the DSP Challenge competition: they plan to continue their work, and have been authorized to pursue the restoration of color film. Faculty adviser to the Singapore team, Dr. Chong Man Nang, leads the parallel processing group at NTU's School of Applied Science. His research interests encompass parallel and distributed processing, scientific visualization, and multimedia signal processing.
The finalists from the University of Maryland used TI's quad-TMS320C40 Parallel Processing Development System (PPDS) tool to design and implement a real-time system that compresses video signals for transmission over narrowband communication channels such as existing telephone lines. The system uses the DSPs to decompose video signals into discrete frequency bands, identifying bands that carry critical information and discarding bands that carrying non-essential data. Developed by students Hamid Jafarkani, Ruplu Bhattacharya, and Jerome Johnson, the system currently reduces video signals by a compression ratio of up to 100.
While many real-time video compression systems currently use custom-designed VLSI technology to solve specific applications, such systems are expensive and time consuming to design. Furthermore, custom-designed hardware is algorithm specific and cannot be reprogrammed to accommodate new algorithms or standards.
In contrast, the system developed by the Maryland team uses TI DSPs to provide fast real-time processing of large data sets, and to implement a fully programmable video compression system that can handle a variety of algorithms and applications.
Nariman Farvardin, faculty adviser to the Maryland team, is chair of the school's department of electrical engineering. Farvardin has previous experience with the Ecole Nationale Superieure des telecommunications in Paris, AT&T Bell Laboratories, and the National Institute of Standards and Technology.
Finalists Xavier Gilles, Sylvain Marques, and Pierre-Henry Dezaux, undergraduate engineering students at the Ecole Fran-caise d'Electronique et d'Informatique (EFREI) outside Paris, developed a multiprocessor DSP system for advanced Doppler radar tracking. Doppler weather radar, for example, is used to detect tornadoes, wind shear, and other dangerous weather conditions.
The team's DSP-based Doppler front end will be used in a new generation of small, light, and programmable tracking radar called AXIR. Because the system is based on conventional DSPs, the AXIR radar system will also cost far less than today's radar systems, making it affordable for small civilian airports and air clubs.
Equipped with four 16-bit 'C50 DSPs, the team's design provides computing power of more than 100 MIPS for sampling multiple antenna channels and executing pulse-compression correlation and adaptive Doppler filtering algorithm code.
Louis M. Fourdan, adviser to the EFREI team, received his degree from Ecole Superierure d'Electricite, and has more than 35 years experience in Doppler radar research and design. Since 1961, he has served as a radar designer for Thomson CSF while conducting DSP, radar, and antenna courses at several engineering schools in France and Turkey.
TI's TMS320 University Program