Abstract
BACKGROUND
The identification of alcoholism is of prime importance because of its adverse effects on the central nervous system. Moreover, people suffering from alcoholism are susceptible to various health problems such as cardiomyopathy, immune system disorder, high blood pressure, cirrhosis, brain anomalies, and heart problems.
NEW METHOD
This study presents a novel approach, based on Fourier theory, known as Fourier decomposition method (FDM) for automatic identification of alcoholism using electroencephalogram (EEG) signals. The FDM approach is employed to decompose EEG signals into a set of desired orthogonal components, commonly referred as Fourier intrinsic band functions (FIBFs), obtained by dividing the complete bandwidth of EEG signal under analysis into equal frequency bands. Time-domain features such as Hjorth parameters, kurtosis, inter-quartile range, and median frequency are extracted from FIBFs. To reduce the complexity, Kruskal-Wallis (KW) statistical test, is performed to adopt the most significant features.
RESULTS
Simulation results are obtained using different classification methods, namely, k-nearest neighbor (kNN), support vector machine (SVM), and linear discriminant analysis (LDA). The proposed approach with the SVM classifier using radial basis function provides average accuracy of 99.98%, sensitivity of 99.99% and specificity of 99.97%. Performance is also tested in the presence of noise.
COMPARISON WITH EXISTING METHOD(S)
Classification results highlight the superior performance of our method in comparison to existing works.
CONCLUSIONS
The proposed scheme provides an efficient approach and can be employed in real-time alcoholism detection.
Collapse