Experimental Evaluation of Steady State Visual Evoked Potentials for Brain Machine Interface

Abstract

This paper presents a signal processing technique to recognize human mental states from steady state visual evoked potentials. In this method, multiple band pass filter is applied to the electroencephalographic signals in order to extract feature points. A neural network classifier is used to recognize the type of LED stimulus that the user is gazing at in real-time. Three experiments were conducted on three participants to evaluate the classification performance, recognition speed, and information transfer capability of the proposed system. In each of these experiments, participants gazed at three types of LEDs flickering at different frequencies. Experimental results showed that the proposed method has achieved 80% of the mean of the correct recognition rate, and it was found that the average of the recognition time was 4.74, s. Although the recognition time was about 5, s, another online classification experiment shows that the average number of decisions per minutes 6.8, and the correct recognition rate was 66.6%.