Penerapan Convolutional Neural Network dalam Deteksi Autism Spectrum Disorder pada Anak Melalui Identifikasi Tulisan Tangan

Abstract

Autism Spectrum Disorder (ASD) is defined by difficulties in communication, social interactions, and behavioral patterns. Children with ASD often face difficulties in fine motor activities, including handwriting, which frequently results in handwriting of lower quality compared to children with typical development. These challenges are often indicative of impairments in motor coordination and sensory perception. Early detection is crucial to providing appropriate interventions, which can significantly enhance children’s development in social, emotional, and academic aspects. This study aims to detect ASD through handwriting identification using Convolutional Neural Networks (CNN). The CNN architecture used is Xception, leveraging depthwise separable convolutions to enhance the efficiency and accuracy of image data processing. The dataset consists of 500 handwritten image samples from children, categorized into four classes: normal, mild ASD, moderate ASD, and severe ASD. All data undergoes preprocessing steps, including normalization and resizing, to ensure data consistency. The CNN method achieved an accuracy rate of 94% on the training data.