One Multimodal Headset to Rule Them All
Yesterday, Kopin a provider of small, lightweight, power-efficent liquid crystal displays, unveiled its latest offering, Golden-i, a speech-enabled Bluetooth headset with a view screen that rides just below the eye line in a user’s peripheral vision.
The device, targeted at the industrial and military spaces for the time being, weighs only 3 ounces. It allows users to remotely control an array of Bluetooth devices, including cellular phones, PCs, networks, and wireless systems, through a combination of voice commands and visual prompts.
Nuance Communications' VoCon 3200 is the voice-recognition engine powering the Golden-i. Kopin officials claim that the device can achieve out-of-the-box recognition rates of “easily over 90 percent proficiency.”
“You know how many times we’ve put this on people in the last two weeks who have heavy dialects and say ‘Speech never works for me! I got speech in my Mercedes. I got this, I got that, and it never works for me!’ But within a couple commands its hearing what they’re saying and within a dozen commands it’s getting really accurate,” says Jeff Jacobsen, senior advisor to Kopin’s chief executive officer and program manager of Golden-i.
“Instead of being like Dragon or Vista, where you have to spend half an hour or 45 minutes training it, we control the horizontal and the vertical. We give you a palette of things to say,” he adds.
The device takes a multimodal approach, directing users by giving them 15 or 16 on-screen icons from which to choose. This functions as a kind of prompter that directs users, letting them know what they can say, which then allows the recognizer to focus more attentively around certain commands.
Because of its target markets, noise cancellation is also an important feature of the device. Jacobsen claims that with the cancellation enabled, the device can operate in environments with an excess of 100 decibels of background noise with the user making utterances at normal volume. “You can’t even hear yourself talking. Your mouth is moving. You think you’re saying words, and the system still understands you,” he explains.
Over the next six months, Kopin plans to do extensive field testing on the Golden-i. Results from those tests will be incorporated as improvements into a version that is expected to be ready for market in 2010.