Correcting errors with redundancy

Data transmitted through noisy and unstable environments is vulnerable to errors. The data flows we count on daily—streaming Star Wars, Skype-ing a child at college, saving a photo to disk—are protected by error correction protocols that add extra bits to data that can be used to recover the message at its destination.

Error correction has been part of communications for decades, says Zhiyuan Yan, associate professor of electrical and computer engineering. Today’s wireless networks rely on methods introduced in the 1990s.

Yan is principal investigator on a $300,000 National Science Foundation project to bring the error protection protocols of the next generation, known as polar codes, into the mainstream.

“Transmitting and receiving bits is not a perfect process,” Yan says. Polar codes, first theorized in 2009, are a promising method to maximize channel bandwidths while reducing the computing resources required to decode on the fly.

“Polar codes are about the integrity and reliability of the data,” he says. During our telephone interview, the analog sound waves of my voice were encoded by my cellphone into digital bits, which flowed through a cell tower to a wired network, to Lehigh’s telephone system, and then were decoded back to sound in Yan’s earpiece. A routine process, yet wavering wireless signals and network congestion can result in errors, Yan says.

Error correction introduces redundancy to correct errors. An elementary (and highly inefficient) method is to repeat data. A binary message, 1011, could be sent as 111-000-111-111. If interference scrambles it to 111-010-110-111, the method would understand the inconsistent triads 010 and 110 as 0 and 1, and the message would be received correctly.

Yan will work on “a multidimensional approach to the two big challenges in the field.” He intends to harness the codes’ efficiency on nearly infinite blocks of data to apply to smaller blocks sent in practice. He will also focus on making hardware implementations of decoding algorithms faster and more reliable.

“A lot of applications require short delays in decoding,” he says. “The mouth-to-ear delay in a phone conversation needs to be less than a fraction of a second. Otherwise it destroys the rhythm of the conversation.”

Other uses, like writing data to a drive, privilege reliability. “You won’t mind if you wait an extra second to access your file,” Yan says, “but you’ll have a big problem if that file is corrupted.”

Story by Robert W. Fisher ’79