Overheard/Underheard

_{The speech-related developments spotlighted in this column don’t quite warrant a full news story, but they’re still too eccentric for us to pass up.}

>>> In March, we discussed End War, a video game in which players command their digital army by voice. Now voice can be used to control a real warplane. The F-35 Lightning II, developed by Lockheed Martin, Northrop Grumman, and BAE Systems, is the U.S. Air Force’s newest toy, a single-engine strike aircraft set to debut in 2009. It’s also the first U.S. fighter to feature voice controls. It uses SRI International DynaSpeak, speaker-independent speech recognition software designed specifically for noisy environments like a plane’s cockpit.

The speech controls were implemented to streamline the pilot’s controls so he can focus on other more important matters, such as incoming surface-to-air missiles or large mountains. Pilot utterances pipe through the oxygen mask microphone.

Currently, scientists are collecting data using recordings from a flight simulator. Their main efforts right now focus on phrases that score particularly low in recognition, and they are tinkering with sensitivity to minimize the error rate, which needs to be less than 2 percent. Adacel Systems will integrate the voice recognition system into the Lightning II. If successful, expect uptake in other planes, such as the F-22 Raptor.

>>> Last month, we ran a story on Ambient’s Audeo, a subvocal speech system designed for people with disabilities. It turns out that NASA has conducted research in this area for the past nine years. Through its Extension of the Human Senses program, a research group is reportedly "developing alternative methods for human-machine interaction as applied to device control and human performance augmentation." In an interview for The Future of Things Web site, Chuck Jorgensen, chief scientist for neuroengineering at NASA’s Ames Research Center, predicted that the technology might be put into commercial applications within two to four years. Currently, the biggest constraints involve limited resources and attaining better recognition for vowels and consonants.

Additionally, Jorgensen states that the Defense Advanced Research Projects Agency (DARPA) has already used the technology in a recent program. He also says that NASA subcontracts for occasional military applications. Jorgensen didn’t go into detail, either vocally or subvocally.