Identification of Single and Multiple Ocular Peaks in EEG Signal Using Adaptive Thresholding Technique

Electroencephalography (EEG) is widely utilized non-invasive method for studying cerebral activities emanating within brain cortex of the order of microvolts. One of the major problems encountered during visual EEG analysis is the presence of artifacts arising from the subject. Ocular artifact (OA) is one of the major sources of unavoidable artifacts that are of high magnitude, thereby making information retrieval a troublesome task. The work discusses the different methodology to be followed for OA identification. This paper implements and compares various statistical and time warping distance techniques which are mostly used for estimating feature distance as well as clustering application for identification of ocular artifacts. Apart from this, the paper proposes a robust adaptive thresholding technique for precise identification of ocular epochs. Along with precision and adaptive characteristics, computational time is also considered as significant factor in the study. The experimental results exhibited a notable performance of 98.4% and 91.68% at an optimal threshold for kurtosis and Dynamic time warping (DTW) respectively but both did not give consistent results for varying datasets (in terms of adaptive and computational time). However, proposed adaptive thresholding technique gave noteworthy results in identifying single, bidirectional as well as multiple ocular peaks surpassing both kurtosis and DTW in all terms.