Making Machines and People See Eye to Eye

Patterns exist everywhere we look, and we recognize them without even thinking

6 October 2008

Patterns exist everywhere we look, and we recognize them without even thinking. A friend’s gait is familiar from a hundred meters or more away; we worry when the pattern of our children’s noise is broken by silence; and we can recognize a TV show by its visual style even though we’ve never watched that scene before.

But we often need machines to recognize patterns for us. “Humans excel in recognizing small changes; machines can recognize massive amounts of data quickly. They go hand in hand,” says Rangachar Kasturi, general cochair of the International Conference on Pattern Recognition, to be held from 8 to 11 December in Tampa, Fla.

“The machines’ goal is to assist humans in their decision-making capabilities,” he adds.

The field of pattern recognition is growing rapidly; more than 1000 papers from 47 countries will be presented at the conference. In addition, a record 1200 attendees are expected, from almost 40 countries. “There’ll be a good mix of industrial and academic people because the field is very application oriented,” Kasturi says.

The meeting draws visitors from around the world because it’s also sponsored by the International Association for Pattern Recognition (IAPR), an umbrella organization of nearly 40 pattern recognition socities around the world. The biennial conference has been held in many countries over the past 35 years. When held in the United States, its technical sponsor is the IEEE Computer Society, one of IAPR’s 40 member societies, of which Kasturi is president.

EVERYDAY SECURITY As technology gets cheaper, more powerful, and more compact, pattern recognition moves into even inexpensive everyday devices, such as digital pocket cameras that recognize and focus on faces, or even refuse to trigger until a subject smiles. Security systems that recognize specific faces have been in the news, as have systems that recognize voices and can transcribe what they say.

Of the two new technical tracks at the conference, biometrics—automated methods for identifying people—has security implications. The other, bioinformatics, or the analysis of large amounts of biological data, deals more with advances in medical and biological technology, such as genomic analysis. This year’s conference will also have a workshop on observing and analyzing animal and insect behavior.

“Pattern-recognition technology often gets its ideas from biological systems,"  Kasturi notes. “Computer vision has been significantly influenced by the psychophysics of human vision. But most algorithms are based on what machines can do rather than what humans do.”

Automated pattern recognition has become very important, now that the balance has shifted from text to visual information, Kasturi points out. Databases are stuffed with visual material that needs indexing as much as text-based files do. But despite at least 20 years of work to identify content based on texture, color, and shape, computers’ ability to recognize and index visual material is still limited and will be an important topic of discussion at the conference, according to Kasturi.

For more information about the conference or to register, visit