A simple introduction to Brain Computer Interface

Original article was published on Deep Learning on Medium

A simple introduction to Brain Computer Interface

Have you ever thought of switching on your television or fan while lazily sitting in your chair and doing nothing but thinking about it? Have you ever wished people/computer understood what you were thinking without even saying or giving a direct command?

Yep it is possible. And you don’t need a magic wand for doing so, instead technology can help you understand and interact with itself better. The devices which can help you perform this magic are called Brain Computer Interface. We are at the early stage of it’s development but it’s almost here. As the definition goes, it is a device or communication system which lets you control any device of your choice just using brain (i.e. not using any muscle or nerve). That means just using your thoughts you will be able to control things or communicate.

THE INPUT

You will be wondering how this is possible.

Your body is made of different type of building blocks/cells. The main functional building block of your brain is called a neuron (nerve cell). These neurons can create electric current when they are functioning. When you perform any sort of mental activity, these neurons fire electricity. Your complex thoughts and existence is a result of these neurons firing in a complex network. All you need to do is record these electrical activities. The different techniques for recording these activities in real time include electroencephalogram, electrocorticogram and intra-cortical recordings.

Intra-cortical recording is literally poking a small needle electrode or array of electrodes into your brain. This is what Neuralink and Facebook is attempting to perfect. These electrodes will be able to record brain’s electrical activity while being in proximity with the neurons. Neuralink has reduced the size of their electrodes to micro-scale, yet these techniques cannot guarantee a zero neural damage in the immediate future. Intra-cortical recording has the most excellent information, but its adoption in normal population is still questionable. Whereas, the electrocorticogram has electrodes placed outside your brain (in the membrane that protects the brain).

Electroencephalography is the technique where you record these electrical activities from outside the skin. No poking, no surgery, no need to put anything inside your body. So it will seem that EEG is best right? Actually no, but practically yes. The ease of use and non-invasive nature makes it ideal for day to day use. But EEG, being complex and noisy makes it challenging for the researchers and developers to create a BCI.

THE TRANSLATION

As you all know our human body is a conductor of electricity. The electric changes happening at neuronal level are passed to the scalp surface from where we record EEG. But as this electrical field is passed on the skin, volume conduction happens and data gets smudged. The electrical resistance of skull only adds to the problem. It’s been reported that one single EEG electrode can record changes from 6cm² to 20cm² of brain surface. This adds to the complexity of the data because now its even more difficult to decode whether the data coming from one electrode is result of one event or a combination of several events happening simultaneously. This is like you are in a noisy club and and trying to make sense of every conversation happening in the room. The neural circuitry in our brain is actually connected to many neurons parallelly. When 100 and more neurons simultaneously show electrical changes, it results in a recordable and detectable EEG.

This also challenges the understanding of what we record in EEG. To a normal human eye, this data may look too noisy. Not much can be understood using simple manual algorithm or statistical tools. There is a complex pattern hidden inside this data, to decipher which we need an algorithm that can understand the complexity with ease. This is where deep learning algorithms come into picture EEG is too complex, unstable and noisy for a simple statistical classifier to handle. But use of deep learning algorithms for classifying this complex EEG data can help the developer give personalized experience to each of the BCI user. Thus, these deep learning algorithms can simplify EEG into simple classes / categories of the corresponding event during which the EEG was recorded.

THE OUTPUT

Using the deep learning algorithm, the computer will know what you think. What will you do with it? Possibilities of how we can use this information are enormous. Let’s take few simple daily life examples.

You use your smart phone every day and you are pretty comfortable with the touch screen. When you go back to the keyboard and mouse in the laptop, don’t you wish for a touch screen (I am pretty sure most of you have already switched to tablet or touch screen laptops). Do you have Siri or Google assistant enabled airpods/headsets? If so, how many times have you asked for current weather or time for no reason? Our interactions with the digital devices are changing, they are getting more and more intuitive and immersive. But, what is even better than touch screen, voice and gestures? What is more intuitive than a simple thought? So why not use thoughts to control your smart devices?

Have you seen a prosthetic limb used by an amputee? Most of the prosthetic limbs available in market give its user limited mobility. What if we can give them their ability back? A complete prosthetic device that will do exactly what the user want it to do.

Are you a gamer? What if your game adjust to your likes and dislikes? What if zombie is too scary for you and animation can be modulated according to your preferences for you to feel better? What if the racing game starts getting boring for you and your opponent challenges you more just to make it more interesting?

Are you a VR user? Have you ever felt that the current controller is cumbersome to use. Have you thought of unlimited freedom in the VR and do anything in your digital realm.

Are you a firefighter or a solider? How about a robot that mimics all your movement, but you will be totally safe back at your home, office or bunker?