Recognition of Dynamic Filipino Sign Language using MediaPipe and Long Short-Term Memory

Filipino Sign Language (FSL) is the official sign language of the Philippines. FSL helped develop the communication gap between impaired individuals, however, only few are knowledgeable regarding the FSL. FSL recognition system can aid the communication gap among Deaf, Mute, and others that serves as a tool to convey message. Previous research includes three types of sign language recognition categorized as data gloves-based, vision-based, and hybrid. Existing wearable devices for sign language recognition which are mostly glove-based, and some hybrid approaches is intrusive as there is still a need for the user to wear something. Existing video-based approaches only recognize static sign language, mostly letters and numbers, only focus on the hand shape, and others require specific background setup. Therefore, there is a need for a dynamic FSL recognition system using MediaPipe and Long-Short-Term Memory (LSTM) capable of detecting pose and hand landmarks and recognize Filipino Sign Language. Using the MediaPipe and LSTM algorithm, the system achieved an accuracy of 95% for the actual testing and has 98% model accuracy. Thus, the recognition of dynamic Filipino sign language using MediaPipe and Long Short-Term Memory is concluded to be highly accurate in recognizing dynamic FSL signs.