A Dataset for Emotion Recognition for Iraqi Autism Individuals as a Step Towards EEG-Based Therapeutic Intervention

n this work, Emotion Recognition for Iraqi Autism Individuals (EmoReIQ) is presented, a dataset of Electroencephalogram (EEG) signals recorded during various sessions for Autism Spectrum Disorder (ASD) participants. Since individuals with ASD often have difficulty understanding and expressing their own emotions, which leads to difficulties in social interactions, communication, and overall well-being; therefore, recognizing and understanding emotions is crucial for them during therapy sessions to provide appropriate support and interventions. Developing, being done for the first time in Iraq country, a dataset that is more specific to the cultural and linguistic context of Iraqi ASD individuals will help treat and try to get them to safety. EEG signals from 28 ASD participants were recorded while they were exposed to visual emotion-eliciting stimuli that evoked one of the five emotions (calm, happiness, anger, fear, and sadness) in different experiment sessions. The classification algorithm, Artificial Neural Network (ANN), is applied and analyzed for emotion recognition. EEG signals were recorded using BrainAccess, a portable and wireless kit that allows the use of effective Brain-Computer Interface (BCI) techniques in everyday applications. A dataset construction protocol is proposed, with emotional stimuli specifically designed to evoke the emotional responses of ASD individuals. EEG data preprocessing and analysis framework is developed to select and combine various EEG-based emotional-relevant features efficiently. With the proposed ANN classifier model, the mean accuracy values are 78.86%, 83.32%, and 72.98% for valence, arousal, and dominance respectively. The EmoReIQ dataset is validated and outperforms state-of-art datasets