Abstract:
"Autism Spectrum Disorder (ASD) presents a complex neurodevelopmental challenge
characterized by difficulties in social interaction, communication, and behavior. One of the
challenges faced by young and adult individuals with ASD, is the he lack of capability to regulate their own emotions. Even though psychological treatments and therapy sessions are held for these children, the parents still lack technological support to monitor the improvement of their child in track.
The solution proposed by this project will use facial images of autistic child to determine their facial emotions. A model was built using CNN. It will be used for Feature extraction and pattern recognition of facial images. Aspects such as the number and the type of layers of the model, activation function and choice of kernels were determined based on the need to improve feature extraction and pattern recognition. It will help to build a model capable of effectively recognizing facial images of autistic children.
The core functionality of the application is built using a sequential model by adding stack of layers one on top of each other. The model has a 7.1% of an accuracy. With better hyperparameter tuning techniques, the model can be improved. The concept of EmoGuard is a novel approach on keeping track of the emotional behaviour of autistic children rather than manual record keeping. May the individual be verbal or non-verbal the application, with a little improvement can be used to gain extremely valuable insights. "
