Humans Still Better Than Machines In Voice Recognition: We Can Identify Close Friends With Utterance Of Just Two Words

The human auditory capacity is exceptional when it comes to identifying familiar voices. Young as they are, for instance, newborn babies are already capable of recognizing the voice of their mothers. They can also distinguish the sound of foreign languages.

The ability of babies to recognize voices and sounds has something to do with the sophisticated sensory memory of humans that provides people with the ability to record and recall large amounts of speech and contextual information.

A new research highlights the exceptional ability of humans to distinguish familiar voices. The new study has found that humans do not need to hear lengthy dialogs to identify the voice of people who are close to them.

The study, which was presented at the 18th International Congress of Phonetic Sciences, found evidence that people who have normal hearing only need to hear two words from a close friend or relative to recognize them.

Study researcher Julien Plante-Hébert, from the Université de Montréal in Canada, conducted an experiment that involved playing audio recordings to participants who were asked to identify which of the 10 male voices they heard was familiar to them.

The researcher came up with a series of "voice lineups," the audio equivalent of what the police use for visual identification procedures in which several individuals who share similar traits are presented to a witness.

In Plante-Hébert's study, participants were presented with a lineup of voices that have similar acoustic aspects. Each voice lineup had varying lengths of utterances, with the longest having 18 syllables.

The researcher found that the participants were not able to recognize short utterances - those with one to three syllables - regardless of how familiar they are with the person speaking. However, with utterances of four or more syllables - the length of two words - the success rate was almost perfect for voices that are very familiar.

Plante-Hébert said that this is one of the things that make humans different from machines and AI systems.

"While advanced technologies are able to capture a large amount of speech information, only humans so far are able to recognize familiar voices with almost total accuracy," Plante-Hebert said.

The findings of the study have potential application in legal investigations where witnesses may identify suspects using limited audio information. The findings may also be used in clinical settings. Children with developmental disorders such as those with autism, for example, are known to have difficulty recognizing their mother's voice.

"Examining normal voice-recognition abilities can orient research on the processes involved in sensory memory," Plante-Hébert said.

ⓒ 2024 TECHTIMES.com All rights reserved. Do not reproduce without permission.
Join the Discussion
Real Time Analytics