A biometric authentication system based on electroencephalogram (EEG) signals allows a person to be authenticated using data that is distinctive to that person and has been independently confirmed. Due to the ease with which common biometrics can be faked or replicated, EEG signal is being incorporated into authentication systems. However, there are still open questions about the system's development, particularly the number of electrodes that will be used for signal acquisition. Current EEG hardware, with its 14 inputs, is impractical for clinical use. Building an EEG-based authentication system with two channels is the focus of this study. Signal processing and categorization are being applied to the captured signal. Entropy and wavelet are used to extract the characteristics. The features are then used as inputs in machine learning algorithms like K-Nearest Neighbour, Support Vector Machine, and Artificial Neural Network. Classifiers are ranked according to their True Acceptance Rate (TAR) in seven different paired-channel categories: AF3&AF4, F7&F8, FC6&FC5, T7&T8, P7&P8, and O1&O2. In order to boost classification efficiency, a fusion algorithm is devised. Weighted fusion of the conditional probability is given as the fusion algorithm by combining three classifier models. The results reveal that the best categorization performance can be found in the P7&P8 channels. The proposed fusion approach outperformed all other classification methods with carefully chosen hyperparameters, achieving an accuracy of 99.2% and 99.1% for classification, respectively. This was followed by SVM (98.1% and 97.3%), ANN (87.0% and 89.8%), and KNN (84.5 and 81.5%).
