Text Entry on Smartwatches Using Continuous Gesture Recognition and Word Dictionary.

This work proposes the development of a method for text entry in smartwatches using continuous gesture recognition, Naïve Bayes classifier and a word dictionary. We performed an evaluation with experts to validate the proposed method. To perform text entry, a user inserts characters through simple gestures, based on geometric shapes, thus, a character is drawn by a user using a set of proposed gestures. We use Naïve Bayes classifier to identify the character that user is entering without the user having to draw it completely. Finally, we use a trie as a dictionary of words to predict words that can be written, considering characters already inserted. We also used the relative probability of word usage in the prediction process. The evaluation with experts showed that it is possible to insert phrases in smartwatches using the proposed method and that the words were inserted correctly.