The ‘the cocktail party effect’ solved with AI

Have you ever found yourself at a party with lots of background noise but still been able to follow what another guest is saying to you? It’s not easy, and there’s a phrase for this situation and the hearing challenges it poses: ‘the cocktail party effect’.

People talking together at a cocktail party

But people with normal hearing and good auditory processing are actually pretty good at listening to one person while filtering out other people’s voices.

For a long time the same couldn’t be said of hearing technology. Delivering suitably clear speech in noise is something the top end hearing aid models have only been able to achieve in the last few years. And how about general audio recordings with multiple background voices?

These too have traditionally been impossible to decipher to work out who is saying what.

For many years, electrical engineer Keith McElveen, founder of US company, Wave Sciences, has been pursuing the problem to find an answer. The fundamental problem has been that sounds bounce around a room. However, with the advancement of AI, Keith has finally had the tools to deliver the clarity in audio recordings that the human ear has long been able to achieve in the midst of a party.

Using AI, and some appropriately placed microphones, Keith can pinpoint and screen out all competing sounds based on where they originally came from in a room. It’s new technology that has supported audio evidence which has been central to convictions in a US murder case.

And the technology is even being used by the US military to analyse sonar signals. Wave Sciences has also released a software application that enables government labs to use its learning algorithm to perform audio forensics and acoustic analysis.

Keith hopes that this is just the beginning. He hopes to introduce versions of the product that can be used in voice interfaces for cars, smart speakers, augmented and virtual reality, and – yes – hearing aid devices.