Design and construct a robotic arm for sign language interpretation with a neural network

Hearing impairment is a problem that affects people at different stages of life. In Ecuador, this affection extends to 14.12% of the total population being approximately 62 thousand people with hearing problems. The use of sign language is presented as an alternative to ensure good communication between people with hearing impairment and people without this condition. However, few people know about this language. For this, we developed a robotic arm with human characteristics to work together with convolutional neural networks for the interpretation of gestures through artificial vision, promoting the learning of sign language in a didactic way. The results obtained demonstrate the ability of neural networks to detect patterns and distinguish the different types of signs in this language. In addition, the use of different neural networks allows a comparison of the limitations of the different models. The efficiency of the models, in terms of percentage accuracy, was 99.27% and 98.2% for the AlexNet and GoogleNet models, respectively. In conclusion, the importance of implementing machine learning techniques to improve communication and generate an impact on the environment of people who are unaware of this problem is demonstrated. It is hoped that in the future this technology can be implemented in the educational system to teach this language to children and young people, thus reducing the communication gap from an early stage.

• Autor principal: Jordan Herrera, Dioselyn Anyeline (author)
• Altres autors: Ubilluz Ortega, Christian Nathanaél (author)
• Format: bachelorThesis
• Idioma: eng
• Publicat: 2024
• Matèries: Brazos robóticos; Visión artificial; Redes neuronales convolucionales; Robotic arm; Artificial vision; Convolutional Neural Network
• Accés en línia: http://repositorio.yachaytech.edu.ec/handle/123456789/867