Research On EEG-based Deaf Emotion Recognition Based On Integrated Genetic Firefly Algorithm

Emotion is a complex mental state or process consisting of both external stimuli and internal activities,and plays an important role in interpersonal communication and interactions.Due to the lack of language communication channels for deaf,they may often have emotional cognitive biases in their daily emotional communication.With the development of brain-computer interface technology,the use of Electroencephalogram（EEG）technology can effectively recognize the emotions of deaf.Therefore,this paper constructs a video-induced deaf emotion induction experimental paradigm,collects EEG data of the three types of deaf emotions（Positive,Neutral,and Negative）,and uses the Integrated Firefly Genetic Optimization Algorithm（IGFA）to identify the deaf emotions.The main research contents of this paper are as follows:（1）Designed and completed an emotion induction experiment based on deaf subjects,and constructed an EEG dataset containing three types of emotions of 45 deaf subjects.Emotional induction of the deaf can be achieved by selecting video clips with a single emotion and strong explosive power.By designing EEG preprocessing steps,a purer emotional EEG signal is obtained.First,the disturbance components in the original signal are removed,and then the artifact components are removed by Independent Component Analysis（ICA）,and each type of EEG signal is segmented according to 6 s to expand the emotional samples.（2）An IGFA is proposed to automatically obtain the EEG emotion discriminant feature and the optimal classifier simultaneously to complete the emotion recognition based on EEG.Firstly,by designing an initialization method based on empirical electrode channels,the three components of random selection,the combination of empirical electrode channels and their random subsets together form the initial population,so that the algorithm has the function of inheriting and learning empirical solutions to reduce invalid iterations.Then,the objective function with variable weights is used to make the algorithm effectively balance the accuracy and the number of feature selection at different iteration stages.The firefly protection and movement rules are used to protect the optimal solution of the population in each iteration of the algorithm,which solves the problem that the Firefly Integrated Optimization Algorithm（FIOA）loses the historical optimal solution.Finally,use the subgroup generation and replacement function to generate new fireflies to replace low-bright fireflies in the iterative process of the algorithm,solve the premature problem of the algorithm in FIOA,and improve the accuracy of deaf emotion recognition.（3）A statistical method of equal-weight emotional discriminative brain regions is proposed to automatically obtain the common emotional discriminative brain regions and individual emotional discriminative brain regions of deaf subjects.By using the total number of feature choices for each subject to find the relative discriminative weights for each channel and frequency band,the cumulative relative discriminative weights for all subjects are calculated to obtain the distribution of emotional discriminative brain areas,solving the problem of variability in discriminative weights between subjects.The validity of the proposed method is verified by comparing it with traditional classification methods based on Power Spectral Density（PSD）and Differential Entropy（DE）features of brain topography.