Mexican Sign‐Language Static‐Alphabet Recognition Using 3D Affine Invariants
Guadalupe Carmona‐Arroyo, Homero Vladimir Ríos-Figueroa, Martha Lorena Avendaño‐Garrido · 2021
Communication for hearing impaired people with all the society is very important. Since most members of this community use sign language it is extremely valuable to develop automatized traductors between this language and other spoken languages. This work reports the recognition of Mexican sign-language static- alphabet from 3D data acquired from leap motion and MS Kinect 1 sensors. The features extracted from the data are six 3D affine moment invariants. These features allow the recognition despite changes in position, pose and shape differences between user's hands. The novelty of this research is the use of six 3D affine moments invariants for sign language recognition. The precision obtained in the experiments with the dataset from the leap motion sensor and using linear discriminant analysis was 94%. The precision obtained using data from MS Kinect 1 sensor was 95.6%.