It might be cheaper and more efficent to wear earplugs while speaking to your computer.

The intent is that in a public setting, a person can talk without being heard by anyone else.

Cell phone users could have a computer generated voice talk for them when when they are in a movie theatre and answering their cell phone.

Physicians have to dictate their experiences with patients and usually have to do it at the end of the day because their is no time during the day where they have privacy. They can not share patient information in any form, so they can not dictate their notes when other patients are around.

In engineering, cubicals are only 6' x 8' and that usually means you can hear everything that goes on around you. If you are dictating a manual or generating specs, you are going to be a little bothersome to the other engineers around you. Everyone does it, so everyone lives with it.

On aircraft, business people want to take notes using a dictaphone, but don't out of consideration for the people around them.

In industry, an operator could direct certain operations to be done in extremely noisy environments.

The point isn't that the user can not hear their own voice, the point is that your thoughts can be translated into words and operations that others can not detect.

The study of this is called subvocalization.

I contacted an non-contact EEG monitor manufacturer about this concept and the discrimination between active areas is still poor.

I contacted an EMG manufacturer and this seems to be a very doable project.

Interesting idea, but you will need to expand it to make it work. Your vocal chords are only part of the system. When you speak, your lips, tongue and teeth determine the difference between many sounds that use the exact same vocal chord vibrations. If you only monitor the vocal chords, you will not be able to tell many words apart. You will need to monitor the muscles of the mouth too.

I got a little experience with an EMG system in one of my biology classes. We were measuring the intensity of muscle contractions produced by using various exercise devices. Based on my experience and knowledge I don't think we will be able to distinguish words this way for a long time. The differences in the movement of muscles to produce different sounds are very subtle. They would be hard to distinguish even with carefully placed electrodes in a lab test. In a real world setting like an office it would be even worse. Maybe someday...

By the way, you said you read somewhere that "every thought you have that you mentally vocalize, that you stress your vocal cords accordingly." I'm rather skeptical of this claim. Can you find a reference?

NASA Begins subvocalization ProjectI'm wondering if they came up with the idea independently. I would like to know the date their research started. I had submitted this idea to NASA as part of a contest they do with NASA Tech Briefs

NASA came up with the idea first. I saw an article dated 2004 where they were using EMG to sense muscle activity. I had been trying to use EEG until last year (2006) because I was interested in using neural nets and genetic algorithms to isolate signal patterns associated with individual phonics. At the time non-invasive EEG was widely available; since then I had stumbled across non-invasive EMG arrays, or Surface EMG arrays.

So this one is now old technology. NASA has been doing it for at least 3 years now. This isn't only possible and practical, it's waiting to be released onto the market.

Junk, the link doesn't work. Can you fix it?

Junk, I searched around and found a report about the NASA project. It does monitor the nerves to the mouth and tongue in addition to the vocal cords. So far, it can only distinguish a small number of words. I would expect that they chose very different sounding words to begin with. Distinguishing similar sounding words will be the real test. We will have to wait and see how long it takes till it has a substantial vocabulary.

This might be more useful in the short term as a computer interface as this application would require few words compared to regular speech on a phone.

In my idea, there was no need for speech to text conversion. The only thing the interface was supposed to do is to correlate phonics. As the person talked a dictionary of phonics (so to speak) would be held in software. As a person would talk they would speak into a microphone. The microphone would feed a DSP processor and it would compare the spoken words with the phonic dictionary and create matches.

In parallel, the EMG Array would feed signals to one or more DSP processors and extract unique and differentiable information from the EMG signals, it wouldn't matter much what the signal combinations were, just so long as they can repeatedly within a certain probability identify a particual phonic.

The DSP processed signals from the EMG Array(s) would then create a weighted value in a neural network and associate the phonics identified by the microphone to phonic dictionary comparison.

Over time, the neural network would isolate the dominant signals from the DSP system monitoring the EMG Array to create a relatively high level of probability for producing the correct phonic.

Learning could be transparent, the person carries on their conversations and ignores the interface. Once the probabilities of EMG matches that of the Microphone interface for producing the correct phonics, the user is notified and can begin using the "telepathic" interface by simply setting up a bluetooth connection with a cell phone, or a mini plug connection to a dictaphone, ...

In the link provided above, when the webpage comes up, click on the link of numbers.

Ah, I understand what you're saying. That sounds like a reasonable idea. Nice one junk.