Abstract
By 2050, approximately 2.5 billion people worldwide will experience some degree of hearing loss, with over 700 million requiring hearing rehabilitation. This growing prevalence emphasizes the urgent need to address communication barriers faced by the deaf and hard-of-hearing population, particularly in healthcare settings. In Kazakhstan, the lack of effective tools for interpreting Kazakh Sign Language (KSL) complicates communication between patients and medical professionals. This study aims to improve healthcare accessibility by developing and evaluating dynamic sign gesture recognition models using MediaPipe for preprocessing. A key contribution of the research is the creation of the KMSG11 dataset, which includes health-related vocabulary in KSL. The proposed Long Short-Term Memory (LSTM)-based model was compared with other existing models, all trained and tested on two datasets—KMSG11 and the Argentinian Sign Language dataset (LSA20)—under the same preprocessing conditions. The results demonstrate the potential of the model to enhance communication, diagnosis, and medical services for deaf individuals in Kazakhstan.