Fuzzy Rule-based Alertness State Classification Based on the Optimization of Eeg Rhythm/channel Combinations

This paper presents a method for automatically selecting the optimal EEG rhythm/channel combination capable of classifying the different human alertness states. We considered four alertness states, namely ’engaged’, ’calm’, ’drowsy’, and ’asleep’. Energies associated with the conventional EEG rhythms, δ, θ, α, β and γ, extracted from overlapping segments of the different EEG channels were used as features. The proposed method is a two-stage process. In the first stage, the optimal brain regions, represented by a set of EEG channels, are identified. In the second stage, a fuzzy rule-based alertness classification system (FRBACS) is developed to select the optimal EEG rhythms extracted from the previously selected EEG channels. The IF-THEN rules used in FRBACS are constructed using a novel bi-level differential evolution (DE) based search algorithm. Unlike most of the existing classification methods, the proposed classification approach reveals easy to interpret rules that describe each of the alertness states.