Gesture Recognition applied to Brazilian Sign Language

Gesture recognition became an important research field in the last years, due to the easy access to different vision sensors, the development of computer vision techniques and the necessity of natural cues for human-machine interaction. Among the solutions for gesture recognition, the use of common video cameras provides a non-invasive and easy to implement the method. This raises one problem: how to represent dynamic gestures in a way that computer models can recognize them with a high generalization capability.

The outcome of this project was a novel technique for hand-gesture description, based on salient points. The technique was evaluated for the tasks of gesture recognition and prediction, using different machine learning methods and showed to be competitive with state-of-the-art solutions.

For more information, please refer to:

• https://github.com/pablovin/Codes-and-Datasets/tree/master/Convexity%20Approach