Web-application development for visualizing the analysis results of the unwanted vocalization recognition model

Abstract

Autism spectrum disorder (ASD) is one of the urgent problems of modern society. Counting the number of vocalizations in children with ASD is an important task for diagnosing and monitoring their behavior. Machine learning methods can serve as an effective mechanism for calculating such vocalizations in the presence of a large sample of audio tracks, and modern web technologies can be used to visualize the results obtained. The article is devoted to the development of a web application designed to visualize the results of analysis performed by models that recognize and classify unwanted vocalizations. The toolkit is based on deep learning methods such as convolutional neural networks (CNN) capable of classifying and counting the number of unwanted vocalizations in children with autism in a classroom based on audio recordings. Using the latest web technologies, the application provides interactive dashboards, customizable data visualizations and real-time analytics, which allows users to effectively study patterns, trends and performance indicators of activities conducted with children. The proposed model shows good results in the classification of vocalizations.