Dyslexia recognition by handwriting using Convolutional Neural Networks

Espinoza-Vicuña C.E.; Nuñez-Medrano Y. · Proceedings of the LACCEI international Multi-conference for Engineering, Education and Technology (2023) · DOI: 10.18687/laccei2023.1.1.1346

Resumen

This project focuses on using technological tools to describe and classify children's handwriting behaviors. This in order to help a proper diagnosis of dyslexia. It was necessary to collect all the necessary information on Developmental Dyslexia. Thus, studying the origin, characteristics, diagnosis, treatments, etc. There are several tests to carry out a diagnosis. These are focused on the study of the phonetic behavior of the voice during a reading. In addition, reader-reader tests, among others. Given this, it has been decided to develop a mobile application, which performs reading-writing tests in participants with and without dyslexia. Thus, finding a large number of patterns that can be studied, mainly related to Dysgraphia, Dysortography and Dyscalculia. After collecting the images from the tests carried out, the second part is to find an artificial intelligence model that allows them to be classified. 2 models were used: convolutional model 1 and the Unet model. These models gave promising results. Finally, the need to continue collecting more images of reading-writing tests in patients with dyslexia was demonstrated.

Instituciones: Universidad Nacional de Ingeniería