A primary goal of using EEGs for collecting data is to convert brain signals into control purposes. Our MATLAB application provides the means to view the EEG data and perform simple signal processing functions. However, due to the random nature of brain signals, one might not be able to accurately produce the appropriate control outputs by visually examining the EEG data and locating the differentiating biomarkers. Therefore, in order to achieve the mentioned end goal, it is imperative to migrate beyond simple signal processing and into the realm of machine learning.
Machine learning is a concept in computer science where a smart algorithm is created for users to feed in data and teach it how it is supposed to interpret the data, in the hope that the algorithm can automatically predict or interpret new data that is fed into the system. To put it simply, machine learning is like having a baby robot that you can train, for example, to differentiate good cars versus bad cars. At first you would need to tell the robot which cars are good and which cars are bad, and then the robot will try to understand what are the features that differentiate good cars from bad ones. The more information (or data) you give to the baby robot, the smarter it is (in doing the classification of good versus bad cars) as it matures.
Our learning application provides the users with a platform to perform machine learning. Users can take real-time EEG sessions and label them accordingly as they are saved into the system. Users can then select their preferred learning model, such as SVM or LDA. Our system will, in turn, perform and validate the learning, as well as allow the users to save the trained learning model. Users are then free to test the learning model real-time.