If you've ever called your bank to check an account balance, chances are good that a computerized voice told you how much money you have. Synthesized voices are increasingly common in call centers. For a business, that voice represents a major investment. So how can companies protect their investments and keep virtual voices from being sampled and stolen?
The answer may come from the same technique that's being used to protect MP3s. Engineers at AT&T Labs are designing a system that will insert an audio watermark into computerized voices, adding high-and low-frequency tones that can't be heard by the human ear. When analyzed by another computer, the tones would not only identify the voice as artificial, but would indicate its creator and owner. Lab researcher Mark Beutnagel says that should help address the concerns of companies that want to protect their investments.
But businesses aren't the only ones worrying about voice theft. The ever-increasing quality of synthesized speech threatens to make it possible to simulate voices so perfectly that they could re-create an actual person's voice. Giga Information Group director Elizabeth Herrell says the production of perfect copies that can fool the ear isn't far away. Imagine a tech-savvy criminal using a computerized version of your voice to impersonate you to find out personal information.
Herrell says one solution to that problem could come from biometrics technology, the use of computerized voice maps to identify people. While speech synthesizers may soon be able to fool the ear, she says, existing speech-verification software can identify a voice far more accurately.
We welcome your comments on this topic on our social media channels, or [contact us directly]
with questions about the site.