*1.2.1 EEG person authentication*

Marcel and Millan [11] investigate the use of brain activity for person authentication, using a statistical framework based on Gaussian Mixture Models and Maximum a posteriori model adaptation. Intensive experimental simulations are performed using strict train/test protocols to show the potential of method [11].

Fladby [12] performs an experiment with 12 participants for eight different tasks in three sessions. They extract features from time and frequency domain by analyzing EEG and then the proposed algorithm is applied as dynamic time warping as well as a feature-based distance metric [12].

He [7] proposes an EEG feature hashing approach for person authentication, by extracting the coefficients of the autoregression model from multiple EEG channels, Fast Johnson-Lindenstrauss transform that is based dimension reduction algorithm to hash vectors. For person authentication, a Naive Bayes probabilistic model is applied [7].

Nguyen et al. [13] investigated the person verification based on brain wave features extracted from EEG signals of motor imagery tasks. For each subject, left, right, and best motor imagery tasks were used. As for modeling, the Gaussian mixture model (GMM) and support vector data description (SVDD) methods were used [13].

Nieves and Manian [14] proposed a system use an effective time-frequency-based feature extraction method using the short-time Fourier transform (STFT) or spectrogram. Computed features on the spectrogram were energy, variance, and skewness. These features were used to train a SVM and neural network classifier. Using cross-validation for testing data for person authentication the classifiers are tested. Results using a different number of channels with optimum features presented [14].

Soni et al. [2] "design a system and implement it, so that users set patterns as an unlock pattern to obtain the access's permission. This pattern can be any combination of eye blink, attention and various brain rhythms like Alpha, Beta, Theta and Delta. Provided two-level authentication. First level of which is brain waves. Once the correct pattern of brain signal is provided the system will ask for a pass key as a second level of authentication [2]."

Sjamsudin [15] "investigates the aspects of performance and time-invariance of EEG-based authentication. Two sets of experiments are done to record EEG of different individuals. The system implemented the use of machine learning such as SVM and deep neural network to classify EEG of subjects [16]."
