Algorithm improves wireless communication

Electrical and computer engineering assistant professor Akbar Sayeed and graduate student Ke Liu have developed novel iterative algorithms that significantly simplify signal processing and improve performance of antenna arrays used in wireless communications systems.

It was long thought that each wireless transmission required a separate frequency. Then in 1998, Lucent Technologies’ Bell Labs introduced BLAST (Bell Labs Layered Space-Time) technology. Multiple antenna systems coupled with signal processing allowed multiple transmissions to occupy the same frequency. Employed at a communications base station, antenna arrays for both transmitting and receiving provide increased capacity and reliability in communication between the tower and the user.

In an array of four transmission and four reception antennas, each transmission antenna can send a separate signal. But on the receiving end, all four receiving antennas get a copy of an individual signal.

To separate the four interfering copies at the receivers and balance the signal to noise ratio, BLAST uses a sequential signal processing technique where the first antenna signal is estimated, computed and then applied to the estimation of the second signal, which is computed and applied to the third and so on. Each estimated signal is progressively more stabilized.

Sayeed and Liu identified an opportunity for improvement in this process. The BLAST system’s first signal estimation is based on a very random signal and never takes full advantage of all four signal copies. In addition, if the first signal estimation were wrong, the error would propagate rather than resolve the signal, creating a bottleneck.

By adding an iteration to the process, Sayeed and Liu found they could dramatically improve performance. Their solution takes the fourth signal estimate and applies it in reverse order.

“The advantage in this case is that we are starting off with the strongest, most stabilized estimate off the transmitted signal,” says Sayeed. “Now your initial estimates are much better and you can actually improve the performance by running a reverse feedback loop. It almost sounds like magic but we see improvements in performance like five or six decibels, sometimes more.”

The teams says the technique could be applied to existing systems with relatively few modifications of original equipment.