**2. The proposed system**

### **2.1 Problem statement**

With the rapid development of technology, large institutions and government institutions, which have sensitive information and also applications, need systems with high security and reliable authentication way that is hard to/or possible to copy or manipulate brain wave biometric has these proprieties, in authentication it is very important that people accept the system (acceptability). With this in mind it is safe to say that a noninvasive method of capturing brain wave signals is the best way for biometric acquisition as for securing these biometrics.

### **2.2 EEG dataset**

In this chapter, the winning BCI competition Graz IV2a Dataset (2008) by burner is used [22], which is consists of nine subjects each performing four MI tasks randomly (this process called trial), each task is a class (left hand, right hand, foot, and tongue) corresponding to (1,2,3,4) respectively, the experiment is done by experts, with no feedback. For our system we use only left and right classes [22].

Each subject sat on comfortable armed chair in front of a computer screen at time (*t* = 0) a fixation cross appeared on the screen and a beep tone to alarm the subject, at (*t* = 2) a cue appear for 1.25 s in form of arrow that refers to one of classes (left, right, down, up) to inform the subject of the beginning of MI tasks this last until *t* = 6 the cue disappears and a break is followed [22].

The dataset is divided into two sessions each in separate day, one for training and other for evaluation, each subject performs 48 trial (each class 12 trial) for 6 runs, yielding in total 288 trial [22]. The sampling rate of the signal is 250 Hz, a bandpass filter was applied between 0.5 Hz and 100 Hz. The sensitivity of the amplifier set to 100 μV, to suppress line noise, a 50 HZ notch filter was applied, and artifact trials were marked. The signals were recorded from 25 channels, 22 EEG, 3EOG [22].
