Facial Action Coding and Hybrid Deep Learning Architectures for Autism Detection

Hereditary Autism Spectrum Disorder (ASD) is a neuron disorder that affects person's ability for communication, interaction, and also behaviors. Diagnostics of autism are available throughout all stages life, from infancy through adolescence adulthood. Facial Emotions detection considered to be the most parameter Autismdisorders among different categories people. Propelled with machine deep learning algorithms, using facial emotions has reached new dimension even been as precautionary warning system caregivers. Since limited only seven expressions, ASD needs improvisation in terms accurate diagnosis. In this paper, we empirically relate Action Coding Systems (FACS) which features extracted by FACS systems. For feature extraction, DEEPFACENET uses integrated Convolutional Neural Network (FACS-CNN) hybrid Deep Learning LSTM (Long Short-Term Memory) classification spectrum disorders (ASD). The experimentation carried out AFFECTNET databases validated Kaggle Autistic datasets (KAFD-2020). Multi-Layer Perceptron (48.67%), neural networks (67.75%), Long ShortTerm Memory (71.56), suggested model showed considerable increase recognition rate (92%), proposed prove its superiority detecting autistic children effectively.