A Comparative Investigation of Wavelet Families for Analysis of EEG Signals Related to Artists and Nonartists During Visual Perception, Mental Imagery, and Rest

Abstract

Differences between the multichannel EEG signals of artists and nonartists were investigated using wavelet coefficients during visual perception and mental imagery tasks and at rest. The wavelet coefficients were calculated using wavelet functions such as Daubechies (order 1–10), Coiflets (order 1–5), and biorthogonal (order 2.4). Each of the calculated approximation and detail coefficients and their averages, standard deviations, and their energies were separately used for discriminating the two groups. The Davies-Bouldin Index was used for evaluation of the feature space quality. We found that the two groups are discriminable using the wavelet coefficients calculated by all of the studied wavelet functions. It was also observed that level of decomposition does not contribute significantly to discriminability. In addition, we observed no considerable difference between approximation and detail coefficients for discriminating the two groups. It was also found that a distinguishing coefficient may exist among the wavelet coefficients, which can discriminate the two groups despite electrode placement. However, separating the two groups is dependant on channel selection when using the energy, average, and standard deviation of the wavelet coefficients. Finally, the two groups were classified by selected wavelet coefficients and a neural gas classifier. The average classification accuracy was 100% for classification of the two groups in the at-rest condition.