BRAINS OUT WITH BCI
when compared with the other signals and they need to go through a considerable amount of filtering to dig out the information.
Source- api. theweek. com
Thanks to modern technology, mind controlled devices have become a fascinating trend among gamers and doctors. The story behind those gadgets is fascinating. Brain Computer Interfaces( BCI) are still not much familiar even with the technical community around the globe. But they are the key elements behind the success of this story of mind control.
A BCI is a bridge of communication between the human brain and an external device. BCI is widely used to analyze brain waves and use them to identify human behavioural patterns, neuromuscular abilities. Rehabilitating disabled patients can be identified as one of the significant contributions of BCI. Modern gaming systems which use BCI may be a familiar product that may concern you. One may come into a conclusion that BCI is just absorbing our thoughts and analyze them to do some tricks. But the framework behind the BCI is far more complex yet fundamentally raised. A BCI is basically composed of three key element areas namely Data Acquisition, Feature Extraction and Classification. It is an evolving machine intelligence interface which learns and improves with training.
Signal Acquisition is usually carried out by placing electrodes on the scalp of the brain. Ten to twenty electrodes system is an standard international method used in electrode placement. These electrodes measure the voltage fluctuations created by the current flowing throughout the neurons of the brain. Such Voltage patterns contain a huge amount of hidden information related to the neuromuscular functionality of our brain. The waveforms are masked with heavy noise. So, one needs a lot of filtering and preprocessing to dig out the brain wave patterns.
There are several types of brain wave signals used in BCI. Electromyography( EMG), Electro-oculography( EOG) and Electroencephalography( EEG) are three main waveforms observed under BCI. In this article, we focus our attention towards EEG signals. Unlike other brain signals, EEG signals contain lots of information on various brain conditions. But of course, details come with a price. These signals are very noisy
Six main EEG wave patterns have been identified and they have been categorized according to their frequency. These different frequency bands give out different characteristics. We’ ll look into them with details. Delta waves are the slowest of the wave patterns being within the range of 4Hz. They are related with the sleeping of adults and also identified normally among babies. Theta waves are within the range of 4Hz to 7Hz. They are seen in the drowsiness of adults and children. They are associated with meditation, relaxation and creative states of the brain. Alpha waves are of the dominant EEG type, being within the range of 7Hz to 14Hz. Alpha waves are related to the closing of the eyes and relaxation. Beta waves lie within the range of 14Hz to 30Hz. They are related evoked with motor behaviour. Gamma waves have a frequency over 30Hz and are known to be associated cross-modal sensory processing and short- term memory matching. And then, there are Mu waves which belong between 8Hz and 12Hz which shows rest-state motor neurons.
As stated before, the waves contain a lot of noise and artifacts. Artifact removal is a challenging task in EEG signal processing. Even the electrical appliances used by day-to-day life cause a disturbance in these wave patterns. And then we have many types of noise, caused by the environmental conditions, magnetic fields and distortions formed by the signal acquisition. Once those signals are filtered and pre-processed they are ready for feature extraction. We cannot find out what the brain is unto by just looking at the waveforms. We need to dig out the specific characteristics that emphasize information related with specific tasks of the brain. Amplitude is one of the basic features of EEG signals. Many features can be extracted from EEG signals in both time domain and frequency domain. Entropy, Energy, Mean, Square integral are some of the commonly used features. Depending on a particular brain class the selected features are compared and plotted with each other. The features that show clear differences and clarify the targeted brain activity are dipped into classification algorithms.
The best part of a brain computer interface is the classification. Here’ s where the machine learning comes into play. Algorithms are imple-
44 University of Peradeniya Gauge Magazine