Brain Computer Interfacing (BCI), or the interface between the human brain and external devices without the need for muscular activity, is a field that has seen tremendous growth over the last two decades. In this survey paper, the basics, signal types, feature extraction techniques, classification techniques, applications, problems, and future trends of BCI will be discussed. In this paper, non- invasive BCI techniques will be discussed, as they have seen tremendous importance. The future trends of BCI techniques will be discussed at the end. Pursuing a computer science degree from one of the top computer science colleges in Maharashtra can help you become a skilled professional in this field.
Background and Principles of BCI
The area of Brain-Computer Interface (BCI), or the interface between the human brain and external devices without the use of muscular activities, is one area that has witnessed tremendous growth over the last two decades. BCI is a field that has witnessed a transition from rehabilitation to gaming, spaces, and assistive technologies. In this paper, a survey on BCI techniques will be presented. Both theoretical and practical aspects of BCI techniques will be presented.II. Background and Principles
The process involved in the BCI processing pipeline includes:
- Signal Acquisition Preprocessing Feature Extraction
- Classification/Regression Application Output
A. Signal Acquisition
The techniques used for signal acquisition are:
- Electroencephalography: This technique records the signals by placing electrodes on the scalp. This technique is the most popular one for non-invasive BCI.
- Magnetoencephalography: This technique records the signals. This technique is expensive and difficult to use.
- Functional Near-Infrared Spectroscopy: This technique records the signals. This technique is portable but has poor time resolution. The most popular technique used for signal acquisition is the electroencephalography technique due to its cost-effectiveness and portability.
III. Feature Extraction and Machine Learning Algorithms
The basic concept on which the BCI works involves the extraction of information from the signals.
A. Signal Preprocessing
The steps involved in signal preprocessing are:
- Removal of artifacts from the signal
- Filters
- Normalisation
The most popular filters used are ICA and band-pass filters.
B. Feature Extraction
The feature extraction techniques used are:
- Time Domain Features: Intensity of the signal
- Frequency Domain Features: Power Spectrum Density
- Time Frequency Analysis: Wavelet transform
C. Classification Algorithms
The machine learning algorithms used in BCI range from traditional to deep learning:
- Support Vector Machines (SVM) – often used in two-class problems.
- Linear Discriminant Analysis (LDA) – an easy and effective solution for most BCI systems based on EEG.
- Convolutional Neural Networks (CNN) – directly learn features from the signal or its transformations.
- Recurrent Networks – often used for dealing with temporal dependencies. Also, hybrid approaches combining different algorithms are being researched.
IV. BCI Paradigms and Control Strategies
Below are the most popular BCI paradigms and control strategies are discussed:
A. Motor Imagery (MI)
The user thinks of moving their limbs, and the neural signals associated with the sensorimotor rhythms are used for control. Examples of controlled devices are wheelchairs and robotic arms.
B. Event-Related Potentials (ERP)
The ERP response, for example, P300, is evoked in response to visual or auditory stimuli. P300 Spell systems are the oldest BCI systems.
C. Steady-State Evoked Potentials (SSEP)
Visual and auditory SSEPs utilise the synchronisation of oscillatory responses to repetitive stimuli.
V.Applications of Brain Computer Interface
-
Assistive and Rehabilitation Systems
The BCI systems assist the motor-disabled individuals to move around and communicate with the world.
B. Human-Computer Interaction (HCI)
The BCI systems allow the individuals to interact with computers and smart devices without the use of their hands.
C. Gaming and Entertainment
Neurogaming is a BCI-based gaming that uses the signals from the brain to influence the games. It is the next level of gaming.
D. Cognitive Monitoring
The BCI systems can be used to monitor the attention levels, workload, and stress levels of the individuals.
VI. Following are the Challenges in Brain-Computer Interfacing
However, there are some challenges that still exist:
- Signal Variability – EEG signals vary in different cases i.e. from person to person and from trial to trial
- Noise and Artifacts – Noise from the environment and the body
- Limited Bandwidth – Limited amount of information that is transmitted from the brain
- User Training – Some BCI systems require a lot of training
- Ethical Issues – Privacy and autonomy
VII. Future Directions
The future trends for BCI systems are:
- Hybrid Models of BCI – Using EEG and NIRS or other biosignals
- Adaptive Learning Systems – Adaptation to individual users in real time
- Wearable Devices – BCI systems can be more accessible to the general public by using these devices.
- Integration with AR/VR – This has helped to improve the immersive experience.
The new signal processing techniques and explainable machine learning can improve the accessibility of the BCI system.
Conclusion
The above survey mainly deals with the current status of BCI technology and the developments that took place in the field of non-invasive BCI, signal processing, and machine learning. You can pursue a B.E. Artificial Intelligence and Machine Learning program to become a skilled professional in this field. The BCI technology has the potential to revolutionise the field of assistive technology, human-computer interaction, cognitive monitoring, and entertainment. There are a few minor issues to be addressed; still, the BCI technology has the potential to revolutionise the above-mentioned fields.