The role of Machine Learning in BCI

Home Human enhancement The role of Machine Learning in BCI

You have a machine extension of yourself in the form of your phone and your computer and all your applications… by far you have more power, more capability than the President of the United States had 30 years ago.

~ Elon Musk
Neuralink CEO

The most popular application of Brain-Computer Interfaces, in the neuroengineering field, is to support people with disabilities. Since they can dialogue with their devices and it allows a considerable improvement in their quality of life.

But not only that, BCI can also be a promising interaction tool for healthy people, with different potential applications in the field of multimedia, virtual reality or video games.

Machine Learning and BCI


Recent technological developments have allowed us to develop artificial intelligence able to make decisions and learn, thanks to Machine Learning mechanisms. However, according to Davide Valeriani, post-doctoral researcher in Brain-Computer Interfaces at the University of Essex, the union between human beings and technology could be more powerful than artificial intelligence itself.

According to the researcher, neurotechnologies would be decisive if decisions were to be taken based on a combination of perception and reasoning. For example, in situations where it would be necessary to decide whether to intervene or not in front of a very blurred image coming from a security camera.

But the biggest challenge for the BCI is that not everyone has the same brain. Using Machine Learning, for each new session, the BCI must learn from the user’s brain by adapting to it in order to properly classify its thoughts. The time taken to do this is known as the calibration time.

Machine Learning is used to create forecasts based on the example input properties, known as training data or training sets. If there is an underlying model containing data properties, a Machine Learning algorithm can build a model based on the available training data as close as possible to the original model. That way, the ML algorithm should be able to predict the class of new input samples correctly, known as test sets.

According to Cortex’s Machine Learning Consultant, Boris Reuderink: “One of the biggest problems in brain-computer interfaces is that the brain signals are weak and very variable. This is why it is difficult to train a classifier, and use it the next day, let alone use it on a different subject. ”

Machine learning approach to BCI. Credits:

Once the data is detected, it can be used for many applications. For example, the subject could use the BCI to control a mouse by imagined movement. A recurring problem is that you must use the data received from the person in the most efficient possible way, but at the same time it is very important to keep in mind that the BCI can make mistakes (for example, the computer may think that the subject has imagined the movement of the left hand, while that of the right hand was imagined). ”

Whatever the BCI technique used (non-invasive, semi-invasive or invasive), the human brain and the machine will work in a symbiotic way with each other. An AI layer based on Machine Learning mechanisms could lie on it, to try to decrease the number of errors, thus connecting us to a new world with the possibility of being at the same level of IA robots.

Start training the BCIs


Once the signal acquisition phases have been completed, where EEG or EMG electrodes must be used, signal conditioning must be carried out: filtering for the frequency desired, filtering of environmental noise, etc.

At this point we have to think about what do we want the system to do for us. Do we want to detect changes in EEG patterns when thinking about the green color? Or do we know a particular change in our EMG when we move a limb? Then we’ll have to think about what we want the computer to do: run a program or simply print the data? Do we want the computer to immediately recognize that a particular data is relevant?

This is supervised learning: choosing a known classification method, obtaining a lot of labeled data and building a system. Methods such as cross-validation can be used to check if the trained models are actually doing what they should.

It is possible to find different EEG datasets available for the public of Machine Learning on brain-computer interfaces in the following links:


Subscribe to our newsletter

To be updated with all the latest news about Brain-Computer Interfaces, subscribe!

· By entering your e-mail address and confirming the following steps, you will accept our privacy conditions.
· Check you spam folder if you don't see the confirmation mail.
Heidi Garcia
Graduate in Computer Engineering at the University of Florence. Editor-in-chief of the Systems: Informatica e Tecnologia su CuE section in the Close-up Engineering network and executive member of EUROAVIA Pisa. Great passion for aerospace, astronomy and electronics.


Please enter your comment!
Please enter your name here