Automatic sign language recognition based on accelerometry and surface electromyography signals: A study for Colombian sign language
[EN] Hearing impairment is a condition that affects the economy and more than the 5% of the world population. Communication between deaf and non hearing impaired people is difficult due to cultural and technological barriers. In this paper, we developed an Automatic Sign Language Recongnition (ASLR)...
| Autores: | , , , , , , |
|---|---|
| Formato: | artículo |
| Fecha de publicación: | 2022 |
| País: | España |
| Recursos: | Universitat Politècnica de València (UPV) |
| Repositorio: | RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia |
| Idioma: | inglés |
| OAI Identifier: | oai:riunet.upv.es:10251/231201 |
| Acesso em linha: | https://riunet.upv.es/handle/10251/231201 |
| Access Level: | acceso abierto |
| Palavra-chave: | Permutation entropy Sign language recognition Accelerometry Electromyography |
| Resumo: | [EN] Hearing impairment is a condition that affects the economy and more than the 5% of the world population. Communication between deaf and non hearing impaired people is difficult due to cultural and technological barriers. In this paper, we developed an Automatic Sign Language Recongnition (ASLR) system of 12 signs of the Colombian Sign Language based on surface electromyography and accelerometry. Initially, we acquired and segmented the signals using a methodology based on multi-objective optimization. Then, we assessed different signal features such as Permutation Entropy (PE) and Root Mean Square (RMS). Finally, we used a Support Vector Machine to classify the signs and a grid search to select the hyper-parameters. The proposed ASLR system showed a low segmentation error of 5.8% and a classification accuracy of 96.66% using only the RMS. These findings suggest that our methodology is suitable to be transfer into embedded systems due to its low computational cost. |
|---|