A Dataset for Electromyography-Based Dactylology Recognition
André Luiz Satoshi Kawamoto, Diego Siedel Bertolini, Maisa Barreto · 2018
Historically, sign languages have been used as a way of communication by people with hearing loss. However, it is not trivial for other people to adopt sign language, due to its particular grammatical structure, rules, vocabulary, and so on. Recently, commercial wearable devices based on the electromyography technology became available to the wide audience at accessible cost. This work shows the process of creating a data set of electromyographic sign records for the Brazilian Sign Language (Língua Brasileira de Sinais - LIBRAS). The authors believe that, using such data set, it is possible to apply machine learning techniques to automatically recognize and translate Brazilian Sign Language to other languages, and thus improve the communication. To demonstrate the dataset potential, some classification methods were applied, achieving up to 89% of correctness.