Sign identification model of the Colombian Sign Language (CSL) alphabet based on Computational Intelligence

Different methods of sign language recognition have been proposed, some based on devices that must be worn by the person to identify signs, these have the disadvantage of having a noisy output, or constant use deteriorates the elements of the instrument. This article proposes a model that allows recognizing the static signs within the LSC alphabet. The signs are captured using a device with an infrared camera called Leap Motion, capable of graphically obtaining the hand that is being displayed in front of it, building a database made up of 38 people. The recognition system is developed using three (3) computational intelligence techniques: Support Vector Machine (SVM), Multilayer Perceptron (MLP) and Random Forests (RF), and a Staking model with the results of individual techniques. Obtaining a recognition percentage of 97.41%, surpassing the SVM by 1.49%, MLP by 3.91% and RF by 4.08%.