A novel functional stock-well transform for ocular artifact removal from EEG

Purposeportable electroencephalogram (EEG) devices have grown in popularity in recent years. However, the artifacts in EEG while capturing occur mostly due to either external or physiological activities. Before clinically relevant data can be extracted, artifacts must be eliminated.MethodsDifferent approaches have been proposed earlier for EEG artifact removal. However, the transform-based method and its modified variants have shown good results. Though large application of wavelet transform is there, still S-Transform is unique as it combines the frequency resolution of the time-frequency space with referenced local phase information. Also, it exhibits a frequency invariant amplitude response, in contrast to the wavelet transform. Further, due to symmetrical property of cosine function, it is used for smooth transition of the signal from one period to another and reduces the leakage effect. As the novelty application of stock-well transform (ST) and its variants are used, though its application for feature extraction and classification problem was performed. Further it is modified as functional S-Transform (FST) for improved result.ResultsFor verification and comparison purpose, different transform techniques are used for experimentation along with variants of ST. The suggested strategy is compared to transform-based methods including the short temporal Fourier transform (STFT), the discrete cosine transforms (DCT), and the discrete wavelet transform (DWT). The evaluating parameters found mean square error (MSE) as1.1554 & mu;V-2, normalized mean square error (NMSE) as 0.8969 & mu;V, relative error (RE) as 0.693, gain in signal-to-artifact ratio (GSAR) as 9,8971dB, signal-to-noise ratio (SNR) as 69.035dB, and correlation coefficient (CC) as 91.55 %.ConclusionsThe efficacy of proposed method is verified, compared and shown in the result section. It is found that the functional S-Transform performs well with measurement of different performance measures as mentioned.