Brain Computer Interface (BCI)
What is Brain Computer Interface?
Brain-Computer Interface is a communicative system that enables its users to interact directly with computers or other external devices through their brain activity. It acquires brain signals, analyses, and translates them into messages or commands that are sent to output devices to carry out desired actions.
The research on BCI started in the 1970s at the University of California and traversed a long way until the recent announcement from Elon Musk about Neuralink: a neurotechnology company developing implantable brain-machine interfaces. The immediate focus of BCI technology is in the fields of biomedical applications and neural rehabilitation to provide communication capabilities to people disabled by neuromuscular disorders such as cerebral palsy, stroke, or spinal cord injury. However, it also has enormous scope in biometric authentication, virtual reality, educational programs, and entertainment applications.
A BCI system has three major consecutive stages: signal acquisition, signal processing, and control interface as shown below.
Types of Brain Computer Interface.
The brain generates a huge amount of neural activities when we think, move or feel, or even blink our eyes. They are in the form of small electric signals that move from neuron to neuron doing the work. Based on the method of brain signal acquisition, BCI is classified into two different types.
- Invasive: electrodes or any other special devices are implanted into the human brain after performing surgery. Though they provide good quality signals, people may not be willing to undergo this kind of surgery unless it is a medical application. In some other cases, a small nanoelectrode array is placed on top of the human brain inside the skull without any major surgery.
- Non-Invasive: devices or electrodes are placed over the scalp to record the neuronal activity of the brain. EEG (electroencephalography), MEG (magnetoencephalography), and fMRI (functional magnetic resonance imaging) are commonly used non-invasive BCIs. In general, these are considered to be the safest and low-cost types of devices.
The brain signals are low-level signals that are amplified to levels suitable for processing. Further, subject to filtering to remove electrical noise such as 60-Hz/50-Hz power line interference and other undesirable movement artifacts. Later, pre-processed signals are digitized and transmitted to a computer for further processing.
In the design of BCI, the selection of relevant features from multiple channels of brain signals is a challenging task. The main aim is to minimize the number of features while maximising the performance of classification. There are many feature extraction methods and few are given below
- Principal Component Analysis (PCA): a linear transformation method in which a set of possibly correlated observations is transformed into a set of uncorrelated variables. Optimal representation of data in terms of minimum mean-square-error.
- Independent Component Analysis (ICA): a powerful and robust tool for artifact removal when artifacts are independent of EEG signals. It splits a set of mixed signals into different sources assuming mutual statistical independence of underlying sources.
- Autoregressive (AR): a model parameter that provides high-frequency resolution for short time segments.
- Wavelet Transform (WT): suitable for non-stationary signals and provides both frequency and time resolution.
- Genetic Algorithm (GA): an optimization procedure to extract an optimal set of relevant features.
The main purpose of classification is to recognize the user’s intentions based on feature vectors extracted from brain signals. Earlier, regression algorithms were extensively used as classifiers, however recently classification algorithms have gained immense popularity. Several classification algorithms are available and few are given below.
- Support Vector Machines (SVM): maximizes the distance between the nearest training samples and the hyperplanes. It is used for both binary and multi-class problems.
- K-Nearest Neighbour Classifier (k-NNC): it uses metric distances between the test feature and their neighbors. Suitable for multiclass problems with low dimensional feature vectors.
- Artificial Neural Networks (ANN): a flexible classifier with multiple architectures available to choose.
- Deep learning methods: they can perform both feature extraction and classification without any human intervention. Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM) architecture are two popularly used in deep learning methods.
The commands from the classification algorithm are used to operate the external device such as cursor control, robotic arm operation, and so forth.
Applications of Brain Computer Interface
Many applications of BCI are currently in use and many more are in research. Few important and promising applications are listed below.
- Medical Applications
Majority of BCI applications is related to healthcare. Few promising areas are mental state monitoring that has contributed to the forecasting and detection of health issues related to seizures, sleep disorders, dyslexia, and autism spectrum disorders. BCI based neuroplasticity can help people suffering from stroke and paralysis. Prosthetic limbs controlled by BCI, also called neural-prosthetic devices, are used to regain the normal functionality of a disabled person.
- Gaming industry
Combining the features of existing games with brain controlling capabilities has opened a huge market for non-medical brain-controlled interfaces. In the Brain Arena video game, players are allowed to collaborate using two BCIs. By imagining left- or right-hand movements they can score goals in the game.
- Education and Self-Regulation
Neurofeedback obtained from BCI is helping to understand the cognitive abilities of each student and helps in personalized teaching. Further, people can learn to self-regulate with the help of cognitive therapeutic approaches using BCI to treat depression and other neuropsychiatric disorders.
- Neuroergonomics and Neuromarketing
BCI in combination with the Internet of Things (IoT) is offering smart environments such as smart homes and transportations. The smart living environment such as music and other lighting conditions can be controlled using the user’s mental state. Further, in the field of intelligent transportation, drivers’ cognitive state is monitored using BCI function. The advertisement industry is using BCI to evaluate the impact of advertisements on users and the role of other cognitive functions that can be used for marketing strategy.
BCI devices in the Market
There are many consumer-grade EEG based BCI devices available for sale. They range from simple single-channel electrodes to 64 channel devices used mainly for gaming, meditation, and education applications. The following list consists of a few prominent devices that have gained popularity in recent years.
- Mind Wave from Neurosky for around $100
- Muse from Interaxon for around $300
- OpenBCI- an open-source platform has various products ranging from $400