Enhancing Keypoint Selection for Hand Model in Recognition of Mexican Sign Language Alphabet

Abstract

The foundation of learning any language lies in its alphabet, which serves as the basis for word formation and expression. This principle also applies to
sign languages, where the manual alphabet is primarily used to spell proper nouns, technical terms, and words without established signs. Technological advancements now enable the tracking and interpretation of human
movements—particularly hand gestures—facilitating the computational modeling of these alphabets. However, achieving efficient real-time performance remains a challenge, requiring the optimization of recognition
models.