Development of real-time sing language transcription system/ Kylle Bryan B. Ausan, Mark Christian T. Carale, Apreal J. De Chavez, Paul Timothy A. Mena, Rebecca R. Sorreda, and Ethan Darrell T. Sta. Maria .--

Thesis (undergraduate)

College of Industrial Technology.-- Bachelor of Engineering Technology major in Electronic Communications Technology: Technological University of the Philippines, 2024.

Includes bibliography:

The study, Development of Real-Time Sign Language Transcription System aims to
present a real-time sign language transcription system capable of translating American Sign
Language (ASL) and Filipino Sign Language (FSL) into English, Filipino, Cebuano and
Ilokano. The system utilizes Raspberry Pi as the CPU of the system, a camera module 2
for capturing sign language gestures, processed using Python and Speeded Up Robust
Features (SURF) technology for gesture recognition. The translated text is displayed on a
7-inch LED touchscreen. The project prototype has a measurement of 5.7 by 8.5 inch for
portability and user-friendliness. It is composed of wireless keyboard, charging port and is
powered by a 10,000 mAh rechargeable battery. This study intends to resolve
communication lapses between individuals who use sign language to interact in different
languages and dialects. The results showed that the sign language transcriber was able to
recognize ASL and FSL and translate best with medium light exposure. The ASLtranslation
has a minimum accuracy of 38%, maximum accuracy of 47%, and an average accuracy of
41%. FSL translation has a minimum accuracy of 49%, maximum accuracy of 56%, and
an average accuracy of 51%. The time delay of translation for ASL has a minimum of 0.87
seconds, maximum of 2.33 seconds and average of 1.55 seconds. FSL time delay has a
minimum of 0.79 seconds, maximum of 2.22 seconds and average of 1.46 seconds. The
system was evaluated by fifty (50) evaluators composed of electronics professionals,
students under the Electronics Engineering Technology Department of Technological
University of the Philippines, engineering students from Adamson University and Signers
using the Confusion Matrix and was rated with an overall mean of 4.31 with a descriptive
rating of “Very Acceptable.” Indicating that the system is beneficial for both signers and
non-signers in terms of communicating.