DEEP NEURAL NETWORK APPROACHES FOR AUTISM DETECTION IN CHILDREN USING VOCAL BIOMARKERS: A SURVEY

Abstract

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by diverse social and communication challenges, often reflected in atypical speech patterns. Vocal biomarkers have thus emerged as a promising, non-invasive avenue for early detection. This survey analyzes approximately 90 peer-reviewed studies published between 2004 and 2024, evaluating deep neural network (DNN)-based methods, particularly Siamese Neural Networks (SNNs), for ASD detection through vocal data. The studies collectively involved sample sizes ranging from 15 to over 1,700 participants, across various age groups from infants to adults. Performance metrics from these studies reported diagnostic accuracies up to 98%, sensitivity reaching 96.7%, and specificities up to 94.2%. The review highlights the effectiveness of SNNs even in limited-data scenarios and outlines challenges such as the lack of standardized vocal features and dataset diversity. It concludes with recommendations for future research to support the development of scalable, real-world solutions for early ASD diagnosis.