Not that the others are good (though I’ve been pretty happy with newcomer Helios, Earthlink’s entry — despite having to spend half an hour persuading them to replace a phone with a new one). Let’s face it: Cellular telephones are a 21st century “convenience” that resembles, in attitude, a collection of feudal empires.

About the best we can say about this technology is that it makes communications easier when it actually works. So I’m pleased to report a breakthrough in cellular technology that should make life with them a little better.

I’m sure you’ve had the following experience: You’re enjoying a meal in a restaurant or a ride in a public conveyance, when suddenly a one-sided conversation grates upon your ears and can’t be ignored.

A few months ago, riding a train to a meeting, I was looking forward to peaceful trip, even trying to catch up on some much-needed rest. As I was nodding off, a woman began a cell phone conversation, displaying an annoying tendency to raise her voice at unexpected moments. It made her impossible to ignore.

You may remember the era of silent movies. Even if, like me, you’re not quite that old you’ve probably seen one or two of them. Its obvious people are talking, yet no sound is coming out.

Now imagine people all around you having cellular-phone conversations, and you can see their lips move yet you hear nothing.

DARPA (Defense Advanced Research Projects Agency) is known as the U.S. Department of Defense’s venture capitalist. It has a project called Advanced Speech Encoding, intended to replace microphones with non-acoustic sensors that detect speech from the speaker’s nerves and muscles rather than from sound.

New Scientist reports that Dr. Rick Brown of Worcester Polytechnic Institute has developed a special collar people can wear that seems to do the trick. In much the same way as magnetic resonance imaging works, the collar detects changes in electrical activity associated with movement of vocal cords. It even allows speech to be detected above loud background noise.

A second approach being pursued at NASA’s Ames lab places special electrode sensors on the neck. These sensors detect very subtle changes in electrical property called “impedance” during speech. A neural network turns these measurements into words.

It can even detect sub-vocal or silent speech.

While DARPA wants this developed for military applications, it clearly would enable speech recording and transmission and permits it for otherwise impossible places, such as loud bars, places with construction going on, and noisy work environments.

There’s no question that this is coming. Prototype systems are expected to be ready for use as early as 2008. Imagine that you’re in a noisy bar or musical event and see an attractive person. You would simply sub-vocalize or speak softly and your words might appear on a small hand-held tablet or even a Wi-Fi-enabled PDA. You can hold a tabular PDA so that the object of your affection could read it, and you could actually tell a good joke and not only be understood, but perhaps elicit a laugh.

A more serious application would enable people with speech defects to lead normal lives.

In terms of productivity, one of the major challenges with using computer speech recognition systems is the variation in people’s voices when they’re sick, congested or just have a dry throat. A second challenge is that such systems are rarely usable in noisy places such as airports or even traveling where the ambient noise varies too much.

With this technology providing input to speech recognition systems such as Dragon NaturallySpeaking 9, the clear leader manufactured by our Transformational Technologies Portfolio holding Nuance (NUAN: NASDAQ), I would expect the final hurdle to ubiquitous adoption of speech systems would be cleared away.

Speaking of hurdles, there is currently one with this technology itself. Words spoken by humans have special acoustic qualities that make us want to listen to them. In the developmental DARPA systems, the words are not spoken but generated by a computer.

Consequently, listeners would hear artificially generated speech that, at least at first, would probably sound something like those annoying tinny recordings still used in some automated speech systems. Eventually, just as is now being done with actors’ voices, the software will be able to synthesize your own voice.

NUAN is in the process of gobbling up the last vestiges of competition in the computer speech recognition market. As I’ve stated before, miniaturization of computer technologies is now approaching a threshold where it will become cost-effective to include speech recognition and synthesis in just about every kind of electronic appliance.

This DARPA technology nails down the final problem, and Nuance has a bright future.

To your profitable future,
Jonathan Kolber
October 25, 2006

Investing in Cellular Technology was originally featured in the Penny Sleuth.