An Information System For Real-Time Mexican Sign Language Translation Using Deep Learning: Promoting Inclusive Higher Education

Inclusive education in Mexico faces structural challenges in addressing the needs of deaf individuals, particularly at the higher education level. In Hidalgo, 41,241 people have hearing disabilities, yet educational continuity and access to accessible technologies remain limited. This project proposes an artificial intelligence algorithm based on convolutional neural networks and Focal Loss for the real-time translation of Mexican Sign Language (LSM) to text, aimed at university environments. A functional web application was developed, trained with a proprietary dataset from a higher education institution, supplemented with public datasets, achieving 99.89% accuracy in recognizing the manual alphabet. The system was evaluated using technical metrics and tests with deaf students and interpreters, yielding a positive correlation (r=0.913) between technical accuracy and perceived inclusiveness. The results show that the tool reduces communication barriers and is viable for institutional implementation as a complement to interpretation services.