Multifractal-Wavelet Based Denoising in the Classification of Healthy and Epileptic Eeg Signals

Abstract

Communicated by Jordi Garcia-OjalvoIdentification of abnormality in Electroencephalogram (EEG) signals is the vast area of research in the neuroscience. Especially, the classification of healthy and epileptic subjects through EEG signals is the crucial problem in the biomedical sciences. Denoising of EEG signals is another important task in signal processing. The noises must be corrected or reduced before the subsequent decision analysis. This paper presents a wavelet-based denoising method for the recovery of EEG signal contaminated by nonstationary noises and investigates the recognition of healthy and epileptic EEG signals by using multifractal measures such as Generalized Fractal Dimensions. The multifractal measures show the significant differences among normal, interictal and epileptic ictal EEGs with denoising by wavelet transform as the pre-processing step. The denoised artifact-free EEG presents a very good improvement in the identification rate of epileptic seizure. The proposed scheme illustrates with high accuracy through the suitable graphical and statistical tools and performs an important role in the epileptic seizure detection.

Time–frequency representation using IEVDHM–HT with application to classification of epileptic EEG signals

“…Crosscorrelation-based multifractal analysis of seizure EEG signals has been performed in [45]. Moreover, methods based on discrete WT (DWT) and generalised fractal dimension (FD) have been suggested for epileptic seizure detection from EEG signals [46][47][48]. A methodology for seizure EEG signal detection has been proposed in [49], where approximate entropy is applied on the components obtained from DWT.…”

Fuzzy Generalized Fractal Dimensions Using Inter-Heartbeat Interval Dynamics in ECG Signals for Age Related Discrimination

“…In 1983, Multifractal theory was developed based on GFD measure [7,8]. The GFD process for noisy signals is discussed in this section [2,3,4,5,7,8,16,19].…”

“…There is an advantage by using the generalized form of fractal dimensions rather than using only some of the dimensions [7,8,10]. The complexity of the time series with nonlinearity has been inspected under different settings by the multi-fractal measure called GFD [2,3,4,5,16,19]. Afterward, GFD has characterized for noisy images to analyze the rate of intricacy [17,18].…”

Audio Magnetotelluric Signal-Noise Identification and Separation Based on Multifractal Spectrum and Matching Pursuit

“…Studies have examined the system's overall characteristics from the local perspective and revealed the essential characteristics of composite signals. 12 Support vector machine (SVM) is a method based on statistical learning theory that can solve the contradiction between the learning and generalization abilities of machine learning by minimizing structural risk. 13,14 Matching pursuit (MP) is the most commonly used sparse decomposition method, and searches the over-complete atom library to realize signal's projection on the optimal atom.…”