Technicalities

WHO SAID THAT? Computerized speech models could be so good in 10
to 15 years that they’ll be able to precisely mimic a
person’s voice, British researchers say. New technology
allows scientists to model a person’s vocal tract after
recording only a sentence of them speaking, which presents a great
leap forward from the old acoustic model of voice impersonation.
Scientists at the British Association Festival of Science recently
told the BBC that modeling a person’s vocal tract based on
analysis of his or her speech would let them create new speech that
sounded exactly like that person — and look like that
person’s voice when digitally analyzed. They warned of the
potential for “vocal terror,” but a more immediate
threat could be vocal crime, to say nothing of the vocal pranks
that could be perpetrated by unscrupulous teenagers, college kids
and political action committees (“Hi, this is a recorded
message from your senator, Bob Klanksdale, confessing to voters
that those claims about kickbacks, bribes and pleasure cruises are
all true!”). It would seem that the technology would be best
used in recordings, because the analysis software couldn’t
model someone’s personality. Still, the new capability, if
perfected, could have a broad impact, not just in crime, terror and
pranksterism, but in areas such as legal prosecutions. The
possibility of accurate voice synthesis, for instance, could throw
doubt on recorded evidence from wiretaps or conversations taped
during clandestine meetings. And, ironically, such a technological
breakthrough might also torpedo the best efforts to date in voice
technology. Although voice recognition isn’t quite a leading
form of biometric authentication, the ability to recreate vocal
tracts might take it out of the running altogether.



About the Author

Kevin McCaney is editor of Defense Systems. Follow him on Twitter: @KevinMcCaney.

Reader Comments

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above