How Unique Is Our Voice in Nature?

[kenny1999]

I know that things like our fingerprint is generally unique and our face is generally unique, unless a twin but chances are rare. How about our voice? Is our unique in nature? Usually how frequent do two persons have the voice indistinguishable by computer or detector? If someone intentionally speaks in a strange way could his voice be seen as a different person?
In addition, what is the parameter that determines the uniqueness of a voice?
Frequency??

 

[Vanadium 50]

If this were strictly true, voice impressions would be impossible.

 

[sysprog]

Not only can the Severe Macaw mimic some human voices accurately enough to fool some humans sometimes; in some instances, it also knows what the words mean.

 

[sophiecentaur]

I believe that voiceprints are acceptable in court but not as the soul form of evidence.
Also, the content and length of the speech clipping will affect reliability and accuracy.

 

[anorlunda]

In addition, what is the parameter that determines the uniqueness of a voice?
Frequency??

Think of two singers (one male one female) singing the same note (same frequency).
Think of a singer singing middle C compared to a trumpet playing middle C.

Obviously, it must be more than frequency. If you think about it, I'm sure you'll be answer your own question.

Edit: Here's a hint. What is the difference between pure and impure tones?

 

[pixel]

This raises doubts about some financial institutions using a caller's voice to verify identity. Maybe it's in conjunction with the caller's phone number.

 

[sysprog]

This raises doubts about some financial institutions using a caller's voice to verify identity. Maybe it's in conjunction with the caller's phone number.

I've never encountered individual voice recognition being used by a financial institution as a sole criterion for identification; I think that, as you suggest, wherever it may be used, it's as an additional safeguard.

 

[anorlunda]

So-called multi-factor identification is widespread for bank software. Each factor may be weak, but together they are strong. The more factors, the better.

If someone mimics your voice or your face or your password, they need to log in from your phone or your laptop also. IP address can also be one of the factors, so can time-of-day when you normally log in, so can typing speed.

 

[sophiecentaur]

The more factors, the better.

Absolutely.
It could actually involve quite a lot of homework on the part of a would be scammer - UNLESS the algorithm for identifying voices could be reverse engineered and a 'machine copy' could be made that would fool the validating software.
Voice recognition would be harder to beat than Mother's Maiden Name or First Pet questions.

 

[anorlunda]

Speaking of multi-factor, I found this in today's news.

Last year, Sergio Caltagirone found himself in a tough spot. While traveling, his phone broke and stopped working completely. With no access to his Google and Microsoft authenticator apps, he lost access to two-factor authentication when he needed it most—when he was logging in from IP addresses not recognized by the 30 to 40 sites he had enrolled.

“I had a whole bunch of sites [that] I had to go through a massively long account restoration process because I lost my 2FA,” said Caltagirone, who is senior VP of threat intelligence at security firm Dragos. “Every time, I had to contact customer service. I had different levels of requirements I had to go through for them to effectively disable 2FA on my account. Some required address verification. [For others,] I had to send a last bill. The number of those I went through was just insane.”

 

[robphy]

Absolutely.
It could actually involve quite a lot of homework on the part of a would be scammer - UNLESS the algorithm for identifying voices could be reverse engineered and a 'machine copy' could be made that would fool the validating software.
Voice recognition would be harder to beat than Mother's Maiden Name or First Pet questions.

Possibly interesting / disturbing ...



these technologies would certainly have methods to quantify how close two voices are to each other.