‘Cocktail party effect’ one step nearer

IMAGINE walking on the street as two women in front of you chat about a new shop, a dog barks behind you, and a police siren wails on the corner.

You turn toward the police car to see what’s happening. When the siren fades out, you realize that the dog is still barking and passers-by are still chatting.

Unknowingly you have found a solution to a problem that has evaded computers for more than half a century: the cocktail party effect.

This is the phenomenon of the brain’s ability to focus one’s auditory attention on a particular stimulus while filtering out a range of other stimuli, as when a party goer focuses on a single conversation in a noisy room. Listeners have the ability to segregate different stimuli into different streams, and subsequently decide which streams are most pertinent to them.

This ability is innate to humans, but alien to computers. Scientists have spent decades trying different approaches to give computers that ability.

Now researchers at the Institute of Automation of the Chinese Academy of Sciences are closer to solving the problem by developing a deep neural network.

They have designed an auditory attention selection model that enables computers to single out useful audio signals from interference signals.

If computers can master the cocktail party effect, it would have a wide range of potential uses. For instance, we could tell our destination to the automatic ticket machine in a noisy subway and buy a ticket immediately. Or you could order a robot in the kitchen to bring you a cup of tea.