Beginners guide to Brain-Computing Interfaces
Imagine being a superhero in those movies, opening portals and doors just with the use of your brain. What if that were something that could actually happen in real life? Well, brain computing interfaces are where it all starts! After a while, we start thinking of all the things we can do while just using our brains. A question we frequently ask ourselves is, can the mind directly communicate with robots, artificial intelligence, and other minds using brain-computer interface (BCI) technology to go beyond what is possible for us as humans?
In this article, we will cover the following:
- What are BCIs?
- Why Does it matter?
- What are BCIs actually capable of?
- How Your Brain Functions Currently and What’s Next
- Different Types of BCIs
- Convolutional Neural Network and BCI
So… what are BCIs??? 🤔
Emerging technology known as brain-computer interfaces (BCIs) links the human brain to a machine or computer (BMIs) and interprets brain signals so the computer can utilise them as input for a job. BCIs are intriguing, but before we go into them, it’s critical to comprehend the brain.
Wait. So why does this even matter?
“The combination of humans and technology might be more powerful than artificial intelligence,” says Davide Valeriani, Post-doctoral Researcher in Brain-Computer Interfaces at the University of Essex. Neurotechnologies, for example, might be utilised to increase perception when we make judgments based on a mix of perception and thinking. This might be useful in scenarios such as viewing a very unclear image from a security camera and deciding whether or not to intervene.” (Link)
What are BCIs actually capable of? 🧠
It all mainly depends if you choose to use the invasive or non-invasive technique. Let us suppose, for the sake of this thought experiment, that healthy people would only employ non-invasive BCIs that do not require surgery. There are now two key technologies in this case: fMRI and EEG. The first requires a large machine, while the second, thanks to consumer headsets like Emotiv and Neurosky, is now available to a wider audience.
BCI, however, can also be a promising tool for engagement for healthy individuals, with a variety of possible uses in the fields of multimedia, virtual reality, and video games, among many other potential uses. The EEG technology is completely safe for the user, but it captures highly noisy data, according to Davide Valeriani. Additionally, research laboratories have so far mostly employed it to study the brain and to suggest novel uses without any follow-up in commercial goods. But that will alter.
The most recent is Musk’s business. In order to monitor signals, its “neural lace” method entails implanting electrodes in the brain. Although it involves surgery, this would enable obtaining brain signals of a far higher quality than EEG. He recently said that in order to prove that humans are superior to artificial intelligence, brain-computer connections are required. (Link)
This technique remains risky! Indeed, we created computers and are fully aware of how they operate and can be “modified.” But we didn’t create our brains, and we still don’t fully understand how they function. much less how to “invade” them in a secure and efficient manner. Although we’ve come a long way, it’s still not enough.
How your brain works now, and what's to come!
Your brain is known to be divided into 2 separate sections, the limbic system and the neocortex.
Our primitive needs, as well as urges associated with survival, including feeding and procreating, are controlled by the limbic system. The most developed part of our brain, the neocortex, is in charge of the logical processes that let us excel in commerce, philosophy, technology, and languages.
The 86 billion nerve cells, or neurons, in the human brain are connected to one another via connections called axons and dendrites. Neurons are active whenever we think, move, or experience anything. The brain does produce a significant quantity of neuronal activity. The task is essentially done by tiny electrical impulses that go from neuron to neuron.
One of the biggest issues with brain-computer interfaces, according to Boris Reuderink, Machine Learning Consultant at Cortext, is how weak and unpredictable brain impulses are. Because of this, it might be challenging to train a classifier and then utilise it the following day, much alone on a new subject. (Link)
Neural Lace is inserted into the skull using a small needle holding the rolled-up mesh. The mesh is then injected, covering the brain as it emerges from the injection site. A lot of emphasis has been paid to artificial intelligence, or machine learning, in the development of BCI applications to address challenging issues in a variety of sectors, particularly those relating to medicine and robotics. Since then, AI/ML has emerged as the most effective technique for BCI systems.
Signal Production
There are 2 ways to produce these brain signals.
- Presenting stimuli (pictures, sounds)
- Hoving them imagine movements
Actively creating signals has the benefit that signal recognition is easier, as you have control over the stimuli; for example, you know when they are offered,” claims Sjoerd Lagarde, Software Engineer at Quintiq. When all you are doing is reading brainwaves from the individual, this is more difficult.
Signal Detection
There are several techniques for identifying brain signals. There are others as well, although EEG and fMRI are the most well-known. EEG and fMRI both assess the electrical activity and blood flow in the brain, respectively. Each of these techniques has its own drawbacks and advantages. While others have a superior spatial resolution, some have a better temporal resolution. Other kinds of measurement methods basically follow the same principle.
Signal Processing
Dealing with brain data has a number of challenges, one of which is that it frequently contains a lot of noise. For instance, when utilising EEG, factors like teeth grinding and eye movements will be seen in the data. It’s necessary to filter out this noise. Actual signals may now be detected using the data. We typically know the types of signals we wish to detect while the person is actively producing them. One such is the so-called event-related potential known as the P300 wave, which manifests when a rare, task-relevant input is given. In your data, this wave will appear as a sizable peak. You may experiment with other machine learning methods.
Signal Transduction
As soon as you find the intriguing signals in your data, you want to put them to use in a way that will be useful to someone. For instance, the individual may utilise the BCI to visualize moving a mouse while using it. One issue you’ll run into in this situation is that although you need to use the information you get from the subject as effectively as you can, you also need to remember that BCIs are vulnerable to error. The current BCIs are quite sluggish and occasionally make mistakes (for example, the computer may believe you thought left-hand movement while, in reality, you imagined right-hand movement).
“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
The different types of BCIs
The 3 main types of BCIs are:
- Non-Invasive
- Semi-Invasive
- Invasive
Invasive approaches require the use of specialised equipment that must be surgically implanted directly into the human brain in order to collect data (brain signals). Devices are placed into the skull on top of the human brain during semi-invasive. Non-invasive technologies are typically regarded as the safest and most affordable kinds. However, because to the blockage of the skull, these devices can only record “weaker” human brain impulses. Electrodes positioned on the scalp are used to detect brain impulses.
EEG (electroencephalography), MEG (magnetoencephalography), and MRT are a few techniques for creating a non-invasive brain-computer interface (magnetic resonance tomography). The most recommended kind of brain-computer interface for research is one that is EEG-based. Control signals, which a computer or robotic equipment can easily understand, are created by processing and decoding EEG data. One of the trickiest steps in creating an effective BCI is the processing and decoding procedure. Particularly, this issue is so challenging that from time to time scientific organisations and different software businesses conduct contests to develop EEG signals categorization for BCI.
Convolutional Neural Network and BCI
An AI neural network based on visual cortex is known as a CNN. In order to reduce classification errors, it has the ability to automatically extract the relevant features from the input data by improving the weight parameters of each filter through forward and backward propagation.
Similar to the visual cortex, the human auditory cortex is organised hierarchically. In a hierarchical system, sensory data is processed in various ways as it moves through the system by a number of different brain areas. Simple cues like colour or direction trigger an action in earlier areas, known as the “primary visual cortex.” Later phases allow for more difficult tasks, such object identification.
Utilizing deep learning techniques has the benefit of requiring less pre-processing because the best parameters are discovered automatically. Feature extraction and classification for CNNs are combined into a single structure and automatically tuned. Additionally, CNN received fNIRS time series data from human individuals. The feature extraction procedure of CNN keeps the temporal information of the time series data collected by fNIRS since the convolution is carried out in a sliding show way.
The non-stationarity of brain signals, however, is one of the main problems in BCI research. This problem makes it challenging for a classifier to identify observable patterns in the signals, leading to subpar classification results.
Hopefully, I have given you a basic understanding of what BCIs are as a whole and that you discovered and learned something new! 🙃
