Detection of Malaysian sign language with single shot detector algorithm/ Nurfarah Idayu Mohamad Fauzi, Shahirah Mohamed Hatim and Zalikha Zulkifli

Sign language is a nonverbal communication that relies on facial expressions, postures, and gestures to facilitate communication between individuals who are deaf or have a hearing impairment. Despite its importance, many people do not recognize or understand sign language, which creates communication barriers for individuals with disabilities. Malaysia faces a lack of interest among its people in learning sign language, which may be attributed to various factors such as limited awareness, resource constraints, or a perception that sign language is not relevant or necessary. To address this issue, the research introduced a simple mobile application for that could potentially increase interest and awareness in sign language and promote greater inclusivity for individuals with disabilities. Single Shot Detector (SSD) algorithm was implemented to perform the Malaysian Sign Language object detection in the application. To facilitate the training of a custom TensorFlow Lite model, the project leveraged the TensorFlow eLite Model Maker library. The outcome of the research indicated a detection accuracy of 75.2%, which is significant as it demonstrates the potential for the developed model to serve as an effective Malaysia Sign Language detector. The framework used in this project can serve as a useful reference for future developers seeking to create similar custom models. Moreover, the promising results of the research indicate the potential for mobile applications utilizing the developed model to significantly enhance communication and inclusivity for individuals with hearing impairments in Malaysia.