EMD-Based Temporal and Spectral Features for the Classification of EEG Signals Using Supervised Learning

Abstract

This paper presents a novel method for feature extraction from electroencephalogram (EEG) signals using empirical mode decomposition (EMD). Its use is motivated by the fact that the EMD gives an effective time-frequency analysis of nonstationary signals. The intrinsic mode functions (IMF) obtained as a result of EMD give the decomposition of a signal according to its frequency components. We present the usage of upto third order temporal moments, and spectral features including spectral centroid, coefficient of variation and the spectral skew of the IMFs for feature extraction from EEG signals. These features are physiologically relevant given that the normal EEG signals have different temporal and spectral centroids, dispersions and symmetries when compared with the pathological EEG signals. The calculated features are fed into the standard support vector machine (SVM) for classification purposes. The performance of the proposed method is studied on a publicly available dataset which is designed to handle various classification problems including the identification of epilepsy patients and detection of seizures. Experiments show that good classification results are obtained using the proposed methodology for the classification of EEG signals. Our proposed method also compares favorably to other state-of-the-art feature extraction methods.

An automated method for sleep staging from EEG signals using normal inverse Gaussian parameters and adaptive boosting

“…Lajnef et al [14] employed various features such as linear prediction error energy, variance, skewness, kurtosis, permutation entropy and multi-class support vector machine on EMG, EOG, and EEG signals for computer-assisted sleep scoring. Riaz et al [15] employed spectral and temporal features extracted in the EMD domain to devise their computerized sleep scoring scheme. Some of the most recent works in the literature performed feature extraction and classification using multivariate linear regression [16], linked component analysis [17], sparse Bayesian learning [18], Bayesian machine learning approaches [19] [20] [21], probabilistic common spatial patterns for multichannel EEG analysis [22], fast nonnegative tensor factorization based on accelerated proximal gradient and low-rank approximation [23], canonical correlation analysis [24] etc.…”

Enhanced Orthogonal Matching Pursuit Algorithm and Its Application in Mechanical Equipment Fault Diagnosis

“…Then, in order to completely reduce the limitations of time-frequency analysis method and better analyze the local time-frequency characteristics of nonstationary or nonlinear signals, Huang et al [4] presented EMD (Empirical Mode Decomposition). It owns the advantages of orthogonality and completeness and has been widely applied in the fields of biomedical engineering, mechanical fault diagnosis, and analysis of seismic signal [5][6][7]. However, it creates problems such as envelope, owe envelope [8], mode mixing [9], and end effects [10].…”

Healthcare data mining from multi-source data

“…• Infrequency: As GHE was offered on a yearly basis, the record sequences were infrequent, compared with time-series healthcare data, such as ECG and body movements collected from wearable sensor devices [88,103,109]. Time-series data from wearable devices were often INTRODUCTION collected in the frequency of Hertz, so the problem was often how to extract more compact data representation to save computational cost [129].…”

“…After all, in our case a person may have more than one record. On the other hand, our problem was also different from the traditional time-series classification problem, which focused on the handling of high frequency series, such as ECG series [129,103,109], because of the infrequency and incompleteness of our GHE records.…”

“…In the literature many transformation algorithms for time-series data were proposed, such as wavelet-based transformation and the Symbolic Aggregate approXimation (SAX) transformation [123]. These methods were usually applied to high-frequency time-series, such as ECG signals recorded in the scale of Hz [109].…”