Autism Detection Using HAAR Cascade Machine Learning Algorithm

Autism Spectrum Disorder It refers to a big spectrum of conditions that influence social interactions, communication skills, and repetitive behaviors. Traditionally, ASD diagnosis relies on behavioural observations, but there is increasing interest in leveraging technology for earlier detection. This project explores using the HAAR Cascade algorithm, typically employed for object detection like facial recognition, to identify different ASD types. We concentrated on four categories: Asperger Syndrome, Childhood Disintegrative Disorder, Pervasive Developmental Disorder-Not Otherwise Specified (PDD-NOS), and Classic Autism are all conditions that fall under the umbrella of developmental disorders. Our approach involved training separate HAAR Cascade models for each type using meticulously labelled images. Positive samples highlighted features associated with each condition, while negative samples included unrelated facial characteristics. The system analyzes new images to classify the type of ASD or indicate no detection if relevant features are absent. Although HAAR Cascade is generally used for simpler tasks, this project aimed to assess its capability in this complex application. The success of our system heavily depended on the quality of the training data and the precision of feature identification by each model. This project is an initial exploration into using HAAR Cascade for ASD detection, suggesting that more advanced techniques, such as deep learning, may be necessary for improved accuracy. Our findings could inform future research, potentially leading to more effective combined methods.