Brain-Computer Interface for Speech

You can read an interesting article about a brain-computer interface to help those who have difficulty speaking (another article about this type of interface here). These devices can be used for those who have amyotrophic lateral sclerosis (ALS) or a stroke of the brain stem. Normally these people are "locked in" meaning they can't move their face or any of their appendages. This can be quite distressing as people are unable to communicate with others. A famous person with ALS, for instance, is Stephen Hawkings. He is currently unable to speak due to the progression of ALS and needs a voice synthesizer.

There are currently several types of brain-computer interfaces that can be done. One is using electrodes implanted inside the brain, the other is non-invasive brain recording via an EEG type device.

There are two choices. One is the patented Neurotrophic Electrode, whereby the electrode tip is implanted 5mm under the surface of the brain and the outer end is attached to amplifiers and FM transmitters located on the skull, under the scalp. No wires or batteries are used. Power is provided by a power induction system similar to your toothbrush holder that charges the toothbrush overnight. This implantation requires major surgery lasting about 10 hours.

The other option is to implant a patented conductive skull screw that does not enter the brain. It records from local field potentials over the surface of the cortex, rather like a very precise EEG (electroencephalogram). These signals can be used to activate a switch and hence provide communication. This option is not actively being pursued at present.

So the basic idea is to translate brain patterns/activity that are directly related to speech. This means identifying the brain mechanisms of speech, using a computer to analyze that brain activity and then efficiently translating that activity into a computer voice synthesizer. So a person would attempt to talk as they normally would, and the their brain activity would be recorded. The computer would then instantaneously make the noise that they wanted to verbalize with a voice synthesizer.

This brings up an interesting point about this project. The brain areas involved in speech are actually quite distributed. There isn't necessarily one single area involved in the production of speech. However, with a device like this, basically you don't really need to know all the complex processes involved in the production speech to be able to reliably reproduce speech in this manner. You don't have to take brain recordings from every area involved in speech in order to get a reliable computer voice synthesizer. You're really outsourcing a lot of tasks to a computer so whole brain reading isn't a necessity. Researchers only have to find brain activity that reliably correlates with particular sounds. It really doesn't matter which area of the brain the activity is actually coming from or the fact that the brain reading is only capturing a tiny piece of the whole puzzle.

It appears that the researchers are taking brain recordings from an area of the brain called the motor cortex for one project. This isn't the only area that researchers have targeted, though (Broca's area is another). The motor cortex is located more on the outer area of the brain and is easier to access especially with non-invasive brain recording. The motor cortex is involved in the movement of parts of the body such as the mouth, the tongue, the lips etc.. So it definitely seems possible that they could take brain readings solely from this area and create a decent voice synthesizer without the need for taking recordings from other brain areas.

I'll admit, I've always been more a fan of non-invasive methods of brain readings/manipulation as I think there more likely to be taken up than undergoing brain surgery to get a device implanted. However these locked in patients don't have much to lose with surgery, so that option seems more viable to them than your average patient.