| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.6 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Atualmente existem bastante dificuldade na comunicação entre a comunidade surda com a restante sociedade. Se esta comunidade tivesse, através de uma aplicação, a possibilidade de comunicação sem que seja necessário aprender língua gestual, um leque de novas oportunidades iria surgir, tanto a nível do mundo profissional, como a nível social. Apesar de já existirem formas de comunicação entre a comunidade surda e ouvinte, estas são, no entanto, bastante complicadas e dispendiosas. É necessário o uso de luvas ou outro tipo de ferramenta externa o que dificulta o acesso e a utilização de tal tecnologia. Esta dissertação tem como objetivo melhorar a inclusão social e a comunicação das pessoas surdas com recurso à tradução de língua gestual para língua oral através de um vídeo, com o mínimo recurso a ferramentas externas. É necessário que o acesso à aplicação seja fácil e acessível em vários dispositivos, para isso a aplicação foi hospedada num website, para que possa ser acedida tanto no computador, como num dispositivo móvel, de forma a prever os movimentos realizados pelo utilizador. No entanto, é necessário primeiramente captar os movimentos e treiná-los com ajuda de modelos de classificação. Finalmente, foram realizados testes de forma a concluir se certas características favoreciam os modelos de classificação utilizados e quais destes contêm uma com maior precisão no acerto das palavras ou frases. Os resultados obtidos foram satisfatórios, o uso da previsão da configuração da mão provou ser eficaz e aumentar a precisão. Este projeto contribui para identificar que o uso de características como a configuração de mão durante a gravação, podem ser significativos, aumentando a precisão, bem como quais os modelos de classificação mais indicados para a previsão de palavras/gestos, ampliando assim a possibilidade e a probabilidade da existência de uma aplicação de fácil acesso sem recurso a ferramentas externas para a integração da comunidade surda.
Currently, there is a lot of difficulty in communication between the deaf community and the rest of society. If this community had, through an application, the possibility of communication without the need to learn sign language, a range of new opportunities would arise, both in the professional world and at the social level. Although there are already forms of communication between the deaf and the hearing community, these are, however, quite complicated and expensive. It is necessary to use gloves or other types of external tools, which makes it difficult to access and use this technology. This dissertation aims to improve the social inclusion and communication of deaf people with the use of sign language to oral language through a video, with minimal use of external tools. The access to the application must be easy and accessible on several devices, for this the application was hosted on a website, which can be accessed both on the computer and on a mobile device, to predict the movements performed by the user. However, it is necessary to first capture the movements and train them with the help of classification models. Finally, tests were carried out to conclude whether certain characteristics favored the classification models used and which of them contains the highest precision. The results obtained were satisfactory, the use of the prediction of the hand configuration proved to be effective and increase the precision. This project helps to identify that the use of characteristics such as the hand configuration during recording, can be significant, increasing the accuracy, as well as which classification models are most suitable for predicting words/gestures, thus expanding the possibility and the probability of the existence of an easily accessible application without recourse to external tools for the integration of the deaf community.
Currently, there is a lot of difficulty in communication between the deaf community and the rest of society. If this community had, through an application, the possibility of communication without the need to learn sign language, a range of new opportunities would arise, both in the professional world and at the social level. Although there are already forms of communication between the deaf and the hearing community, these are, however, quite complicated and expensive. It is necessary to use gloves or other types of external tools, which makes it difficult to access and use this technology. This dissertation aims to improve the social inclusion and communication of deaf people with the use of sign language to oral language through a video, with minimal use of external tools. The access to the application must be easy and accessible on several devices, for this the application was hosted on a website, which can be accessed both on the computer and on a mobile device, to predict the movements performed by the user. However, it is necessary to first capture the movements and train them with the help of classification models. Finally, tests were carried out to conclude whether certain characteristics favored the classification models used and which of them contains the highest precision. The results obtained were satisfactory, the use of the prediction of the hand configuration proved to be effective and increase the precision. This project helps to identify that the use of characteristics such as the hand configuration during recording, can be significant, increasing the accuracy, as well as which classification models are most suitable for predicting words/gestures, thus expanding the possibility and the probability of the existence of an easily accessible application without recourse to external tools for the integration of the deaf community.
Description
Keywords
Automatic translation Video Research Sign Language Classification Models