ERAnet: Emotion Recognition based Assistive learning network for autistic children

Children with Autism Spectrum Disorder (ASD) often face difficulties with traditional learning methods, particularly in understanding emotions, interpreting social cues, and maintaining attention due to hyperactivity. To address these challenges, we propose the Emotion Recognition-based Assistive Learning Network (ERAnet), which consists of three main phases: the ASD Learning Phase, the Emotion Recognition Phase, and the Audio Analysis Phase. In the ASD Learning Phase, facial emotions are detected and translated into emojis that serve as learning cues for the child. During the Emotion Recognition Phase, the child attempts to identify the displayed emotion by matching it to the correct emoji, with up to three attempts allowed. In the Audio Analysis Phase, the child's facial reactions while listening to audio are monitored to compute an emotion score. We thoroughly evaluated the model's performance using standard metrics such as precision, recall, F1-score, and accuracy. The model was also validated on benchmark datasets, achieving an accuracy of 91.45%. Additionally, we tested the model's real-time effectiveness through interactive sessions with autistic children. The results indicate that ERAnet outperforms current state-of-the-art methods.