FIGS-DEAF: an novel implementation of hybrid deep learning algorithm to predict autism spectrum disorders using facial fused gait features

Autism spectrum disorder (A.S.D.) is considered a heterogeneous mental disorder, which is notoriously difficult to identify for a better diagnosis, especially among children. The current diagnosis methodology is purely based on the behavioural observation of symptoms prone to misdiagnosis. Several hybrid methods were explored, which also needs its improvisation in better prediction and diagnosis to move this field towards intelligent and accurate diagnosis. The main objective of this research paper was to develop the new diagnosis software which integrates the novel fuzzy hybrid deep convolutional neural networks and fusion of facial expressions and human gaits based on input video sequences. The algorithm has been
trained and validated with the different video datasets such as Kaggle FER2013 and Karolinska Directed Emotional Faces (KDEF) datasets with real-time test scenarios, and various parameters such as accuracy, recall and F1-score were evaluated. Our proposed deep learning model outperforms another state-of-the-art method with an increase in prediction accuracy up to 30% with maximum accuracy of 95%. The model presented in this paper yields more advantages in terms of prediction time and accuracy also. However, the speech therapists, teachers, caretakers, and parents
can use the software as a technological tool when working with children with A.S.D.