Abstract
This paper describes a method for band-pass filtering of brain waves using the Haar wavelet and of detecting the P300 component of event-related potentials (ERPs). For the visual stimulus of the word-selecting experiments, 5 words were used. The words were selected in a pseudo-random sequence and displayed on a screen for 300-400ms with a cyclic period of. 500-700ms. The target word was randomly applied with a probability of 20%. The subject was instructed to keep a mental count of the number of target words. The P300 components were obtained from the target responses utilizing the following methods: 1) Band-pass filtering of the central frequency of 3Hz, 2) Averaging the single-trial responses to each stimulus 30 times, 3) Subtracting the common components in ERP responses for both target stimuli and nontarget stimuli, and 4) Cutting out the P300 using the time-window of the trapezoidal form. Results show that the P300 component can be used to detect the attention of the subject without any physical action. This P300 component is useful as a human-interface for communication aids.
