The idea: Use voice recognition to authenticate, by speaking aloud a passphrase or a text generated by the system with which you're trying to authenticate.
Examples: Porticus, a startup profiled back in 2007, has an implementation of this technology (VoiceKeyID), available for multiple mobile and embedded platforms.
The good: The phrase used to identify you isn't the important part; it's the voice itself. Also, it can be easily changed; speaking is often faster than typing or performing some other form of recognition; and the solution even works in a hands-free environment. Plus, microphones are now standard-issue hardware.
The bad: As with any technology that exists in a proprietary, third-party implementation, the hard part is getting people to pick up on it.