Resumen

ORTIZ-FARFAN, Nelson  y  CAMARGO-MENDOZA, Jorge E.. Computational Model for Sign Language Recognition in a Colombian Context. TecnoL. [online]. 2020, vol.23, n.48, pp.191-226. ISSN 0123-7799.  https://doi.org/10.22430/22565337.1585.

This document presents the implementation of a Colombian sign language recognition software for deaf people. For this purpose, Machine Learning will be used as the basis of the specific system. Today there is no public repository of images or video that contains these signs or the information necessary to achieve this goal, being one of the main obstacles to undertake the task. For this reason, the construction of a repository was started. Despite the time constraints of the participants, five people carried out the signs in front of a video camera, from which the images that would make up the repository were obtained. Once this was done, the images were used as training data for an optimal computer model that can predict the meaning of a new image presented. We evaluated the performance of the method using classification measures and comparing different models. The measurement known as Accuracy was an important factor in measuring the different models obtained and thus choosing the one most suitable. Results show that it is possible to provide new tools to deaf people to improve communication with others who do not know sign language. Once the best models have been chosen, they are tested with new images, similar to those in the training, where it can be seen that the best model achieves a success rate of around 68 % of the 22 classes used in the system.

Palabras clave : Deaf people; Machine Learning; computational model; sign language.

        · resumen en Español     · texto en Español     · Español ( pdf )