Redes neuronales convolucionales aplicadas a la traducción del lenguaje verbal español al lenguaje de señas Boliviano: Convolutionary Neuronal Networks Applied to the Translation of the Verbal Spanish Language to the Bolivian Sign Language
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Rev. Cien. Tec. In.
Abstract
Se ha utilizado redes neuronales convolucionales para interpretar sonidos emitidos por personas, para posteriormente ser traducidos al lenguaje de señas boliviano, se recurrió a la transformada de Fourier y las escalas de Mel para la creación de patrones de entrenamiento, con diferentes tamaños considerando palabras sueltas y frases del español y una red neuronal convolucional para el reconocimiento. El entrenamiento de la red neuronal considero todos los tamaños de patrones con la finalidad de mejorar el filtrado de la voz capturada antes de aplicar el proceso de reconocimiento. La interpretación de la traducción utilizó el lenguaje dactilológico y el lenguaje de señas boliviano y su representación visual se la realizó a través de animaciones en tercera dimensión. La efectividad de la traducción fue validada a través de un experimento con la participación voluntaria y autorizada de internos del instituto audiológico en la ciudad de Sucre.
It has been used convolutional neural networks to interpret sounds emitted by people, later translated into Bolivian sign language, the Fourier transform and the Mel scales were used to create training patterns, with different sizes considering single words and Spanish phrases and a convolutional neural network for recognition. The training of the neural network considered all the sizes of patterns in order to improve the filtering of the captured voice before applying the recognition process. The interpretation of the translation used the sign language and the Bolivian sign language and its visual representation was realized through animations in third dimension. The effectiveness of the translation was validated through an experiment with the voluntary and authorized participation of inmates of the audiological institute inthe city of Sucre.
It has been used convolutional neural networks to interpret sounds emitted by people, later translated into Bolivian sign language, the Fourier transform and the Mel scales were used to create training patterns, with different sizes considering single words and Spanish phrases and a convolutional neural network for recognition. The training of the neural network considered all the sizes of patterns in order to improve the filtering of the captured voice before applying the recognition process. The interpretation of the translation used the sign language and the Bolivian sign language and its visual representation was realized through animations in third dimension. The effectiveness of the translation was validated through an experiment with the voluntary and authorized participation of inmates of the audiological institute inthe city of Sucre.
Description
Vol. 12, No. 13