The way you breathe could be giving you away. The patterns of inhalations made by people when they speak seem to be unique to each individual – which means an algorithm can be used to identify the speaker, even if they don’t say anything.
The inhalation sounds are turbulence rather than noises made by vocal chords. Their fingerprint is determined by many factors including lung capacity and the shape of the speaker’s oral passageways. The turbulence is referred to as “intervocalic breath sounds”.
“These breath sounds are evident even in poor quality recordings, like telephone conversations,” says Rita Singh at Carnegie Mellon University in Pittsburgh.
Advertisement
She and her team analysed thousands of recordings of news anchors and people being interviewed on TV news programmes. The team found that when analysing inhalations only, their system could identify the correct speaker out of a group of 50 with nearly 75 per cent accuracy.
Who’s there?
The team has also found that it was easy to distinguish speakers who impersonated the voice of another individual – Donald Trump – by noticing differences in their inhalations.
“There’s been years and years of work into speaker identification from their voices,” says Roger K Moore at the University of Sheffield, UK. He says the results of focusing on breath sounds instead is “remarkable”.
Singh has been working with law enforcement on ways to identify people who have made threatening or hoax phone calls. The latter frequently waste emergency services’ time and money.
The hoaxers don’t use their own natural voices, but instead try to sound like someone else who is in panic, says Singh, who has listened to recordings of such calls. “I was amazed to see the range and types of disguise that people are able to instinctively create.” But their breathing signature should still be enough to identify them.
Even if a hoax caller scrambled their voice electronically to avoid being identified, it is possible that enough identifying information could be revealed by their breathing, although Singh has not explored this yet.
“We’re looking more deeply into this as of now – who knows what we’re going to discover,” she says.
Read more: Your voice betrays your personality in a split second
