DOI: 10.14704/nq.2022.20.8.NQ44736

Optimized hybrid channel selection technique with Deep Q network for human emotion classification with EEG signals

S.Lokesh, T.Sreenivasulu Reddy

Abstract

Emotion represents complex psychological disorder whichhappens from response of brain and is considered as an imperative part in day-to-day lives. In recent days, several researches focus on emotion, but to build effectual user interface applications, several techniques are adapted that focus on behavioural representation. Anovel optimization model is presentedto classify emotion withElectroencephalogram (EEG). The selection of optimal channel is done by combining wrapper and entropy techniquewith newly devised optimization technique, namely Horse Fractional Chimp Optimization Algorithm (HFrChOA). Here, the proposed HFrChOA is generated by blending Horse Herd Optimization Algorithm (HOA) and Fractional Chimp Optimization algorithm (FrChOA), which is combination of Fractional Calculus (FC) and Chimp Optimization Algorithm (ChOA). The classification of human emotion is donewith Deep Q Network (DQN). The weights of DQN are optimally tuned with proposed HFrChOA. It noted the emotion recognition using the EEG signals is done by examining the cues of emotions considering time domain features. The proposed HFrChOA-based DQN offered improved competence with high testing accuracy of 91.5%, sensitivity of 89.5%, and specificity of 91.6%.

Keywords

Human emotion classification, Electroencephalogram (EEG) signals, Deep Q network, Entropy, Wrapper feature.

Full Text

References