Performance Enhancement in P300 ERP Single Trial by Machine Learning Adaptive Denoising Mechanism

The P300-based lie detection scheme is yet another and advantageous tactic for unadventurous polygraphy. In the proposed scheme, the raw electroencephalogram (EEG) signals are assimilated from 15 subjects during deception detection. After the assimilation, EEG signals are separated using an independent component analysis. The proposed adaptive denoising approach, extracts three kinds of features from denoised wave to reproduce P300 waveform and identify the P300 components at the Pz electrode. Finally, in order to enhance the performance, four classifiers are used, i.e., support vector machine, linear discriminant analysis, k-nearest neighbor, and back propagation neural network, achieving the accuracy of 74.5%, 79.4%, 97.9%, and 89%, respectively.