Use of Modified S-Transform for EEG Artifact Removal

Stock-well transform (ST) as an effective tool for time-frequency analysis has been utilized for some decades. However, its application on artifact removal is a new direction in biomedical research. In this paper, the authors tried to eliminate the artifact from the brain signal. The brain signal is collected from Mendeley and the PhysioNet database. The ST is modified by considering the orthonormal property of the signal along with the symmetric property. The modified orthonormal S transform (MOST) preserves the non-redundant samples of the signal, similar to application of discrete orthonormal S transform (DOST). Further, the artifacts are removed well, keeping the conjugate symmetry property so that low frequency EEG signal is reconstructed as original signal. For the comparative analysis DWT, DCT, ST, and DOST are also verified, and found that the proposed MOST technique well performs as compared to other transform-based methods with low complexity of $O (N \ log \ N)$ similar to the DFT computation.