Classification of Brain Electroencephalogram Signal based on Adaptive Neural Networks Model for Brain Computer Interfacing

A Brain-Computer Interface (BCI) is a communication system between human and machine (computer) that does not require any human muscular activity. BCI enable human to operate an electronic device only by means of brain activity. Brain signal is captured and translated into command. The most benefit of BCI system is people who is suffering from motor disabilities (paralyze) to control assistive devices such as wheelchairs. A special type of brain signal is called  Electroencephalograms (EEG) is used in BCI system. The EEG signals are the actual carriers of conscious experience. EEG signals are recorded using a set of sensors called electrodes (channels).

(1) The first problem is the difficulty of detecting accurate EEG signals due to existence of artifacts and noise along with signals. Electrical signal from other brain cells is mixed with original EEG signal.  part from artifact problem, weak EEG signal is another problem. The head skull is a main barrier between brain cell and capture device. In this scenario a well-known method called classification in pattern recognition plays an important role in detecting the clean EEG signals. Good classification method can overcome the uncertainty in detection EEG has always been a challenge in BCI system. 

 (2) The second problem is the EEG-action mapping issue. EEG-action mapping issue is to identify which EEG signal relate to action (mapping EEG to human action). One method of EEG-action mapping method is by simulator. Simulator is a kind of tool to stimulate the brain to produce signal. However it requires high focus and extensive train to the human. In this research we try to focus on smart-home action such as controlling television and switching light and indeed, the design of suitable and user friendly simulator for EEG-action mapping task in smart-home activities is a challenge work.

The objectives of the research are as follows:

1. To identify the features and characteristics of neural network satisfy to EEG characteristics.

2. To devise a new classification algorithm that accurate and fast in detecting EEG signal.

2. To devise a new simulator for EEG-brain mapping task specifically for smart-home environment.

The research consists of four main stages, Simulator design and EEG reading phase, pre-processing and feature extraction phase, training the proposed work phase and finally testing and evaluation  phase. Simulator design is the design of simulator for brain stimulus on reflecting the home activities (smart-homes application). Once simulator is ready then the raw EEG reading will be collected from subjects (users). Data processing and features are cleaning the raw EEG and object trials, signal sample, electrodes and etc. In this stage the proposed neural network based classifiers is started. Training with Neural Network models is refining the NN model to produce the trained classifier.

The trained classifier then go through testing and evaluation phase. Test data set will be used in testing phase from new subjects. Performance metric know as recall or precision will be used to measure the performance. Terms such as false negative (FN), true negative (TN), true positive (TP) and false positive (FP) are used.