Construcción de un prototipo electrónico traductor de lenguaje de señas a voz con base en procesamiento de imágenes.

Abstract

The objective of the present research was to construct an electronic prototype for translating sign language into speech based on image processing. This prototype is made up of three fundamental blocks: one for data acquisition, which includes a USB camera; a processing block, composed of a Jetson Nano development board, where the processing of the captured images is carried out; and finally, a block for displaying and reproducing information, composed of a screen and a speaker, respectively. The central block extracts data from the images, specifically from the 21 knuckles of the hand. These data allow the translation of the letters of the Ecuadorian Sign Language alphabet into the standard alphabet, enabling the generation of a wide variety of words by the concatenation of these letters. Finally, these words are presented to the user in text and sound formats. The results of the tests conducted at distances of 30 cm, 60 cm, and 100 cm indicate that the best performance was achieved at a distance of 30 cm, with a detection accuracy of 93.21%. It is concluded that the developed prototype achieves good translation, making it a valuable tool to improve communication for individuals with hearing impairment with those who do not have such disabilities. It is recommended to have adequate lighting in the capture environment and avoid backlight situations because some lighting conditions can make it difficult for the prototype to identify the signs and gestures made by the user.