How We Identify Single Voices in a CrowdS

There are plenty of human abilities that we take for granted, but which are actually insanely complex. Like picking out a single voice buried amongst the noise of a crowded environment, a problem which has troubled scientists for decades. But now they've worked out how we do it—and it could revolutionize speech recognition technology.

The phenomenon—sometimes called the cocktail party effect—allows us to pick out the voice of somebody when all around us is noise. Now, a team of scientists from the University of California, San Francisco has performed experiments on patients undergoing brain surgery to discover how that works. The findings appear in this week's issue of Nature.

During the surgeries, a thin sheet of 256 electrodes was applied to the temporal lobe—the auditory cortex of the brain—of the participants in order to record neuronal activity. Post-surgery, patients were played audio tracks with multiple voices, and asked to identify the words uttered by particular speakers while their brain activity was monitored.

The researchers then used software to reconstruct the brain's activity and assess how it varied when the patients were listening out for different speakers. Amazingly, the neural cortex only seems to respond to a single voice at a time when we're concentrating on making it out, effectively shutting out the rest of the acoustic environment which surrounds us. In other words, selective hearing is very much real—we only hear what we want or need to.

While it's a neat insight, the researchers are also hopeful that it could be a useful tool in assessing hearing impairment and attention deficit disorder. Not just that, they also hope to develop devices for decoding the intentions and thoughts from paralyzed patients that cannot communicate.

And then there's one last, and potentially very lucrative, application: voice recognition. One of the major stumbling blocks with Siri and its brethren is their inability to cope in noisy environments. If scientists can get to the bottom of how the temporal lobe itself filters out extraneous noise, consumer technology could make a huge leap forwards. [Nature]

Image by Brian A Jackson/Shutterstock