Article Preview
TopIntroduction
Electroencephalograph (EEG) is an electrical signal recorded from the brain and very effective to diagnose the diseases related to the brain (Sanei & Chambers, 2007). But the high-quality EEG signals are contaminated by noise. It has been noted that the presence of noise makes an accurate evaluation of EEG signals difficult (Reddy & Narava, 2013). These noises may be generated because of the wrong placement of the electrodes or because of the eyeball, hand, auditory or motor movement (Karthik, Fathima, Rahman, Ahamed, & Ekuakille, 2013). These artifacts are not recognizable visually and called as evoked or event-related potentials (Machado et al., 2010). EEG Signals are having very low SNR and as ERPs mixed with it, are very week signal. To show the feasibility of the investigation of ERPs examination, it is important to have a high SNR estimation of EEG signals. So, the main aim of doing research in the field of EEG signal is to find out a method or technique to extract ERP waveform from the EEG recorded signal with high SNR of EEG (Paulchamy, 2017).
There were a couple of techniques proposed for Electroencephalograph signal handling and have genuine criticalness in specialized utilization (Quiroga, 2006). There is one such method by which, the noise level of EEG signals can be reduced and SNR can be enhanced is, wavelets (Wang et al., 2007). To find and remove those noisy fragments from electroencephalograph signals, discrete wavelet transform (DWT) has been associated (Asaduzzaman et al., 2010). In any case, Wavelet-based systems are very complex. As of now preliminary estimation of peak latency and amplitude of multiple correlated ERP components has additionally done (Ranjbar, Mojtaba, Mikaeili, & Banaraki, 2017). To remove the contamination from the EEG signal in a single trail, filtering can also be a good option but not using a normal filter, as the filter performance based on the statistical properties of the signal to be processed. The solution to this problem will be overcome by adaptive filters (Widrow et al., 1975). Adaptive filters are a nonlinear filter which depends upon the characteristics of the input signal. So, adaptive noise cancellation technique can be a good option for noise removal from signal (Szalai, Haller, & Marthi, 2012). In adaptive noise canceller, there is a need for an adaptive algorithm to update the weights of the filter. The literature review explores the gradient-based algorithm knows as Least Mean Square (LMS) (Górriz et al., 2009), Recursive Least Square (RLS) (Lee & Mathews, 1994) and many variants of it (Widrow, 2013; Diniz, 2008; Khanom & Islam, 2017). Recently EEG signal enhancement is also done by artificial neural network where Kalman filter is used to train it (Yakoubi, Hamdi, & Salah, 2018).