Mobile Sign Language Recognition for Bahasa Indonesia using Convolutional Neural Network
Pujianto Yugopuspito, I Made Murwantara, Jessica Sean · 2018
Hand gestures for speech impaired community have their usage for specific language. In Indonesia, hand gesture has their natural two hands sign and widely accepted usage, BISINDO (Bahasa Indonesia Sign Language). In this paper, we propose to use a mobile application to support people who want to communicate with speech impaired people based on BISINDO. We make use the Convolutional Neural Network method to identify the hand gesture in a real time Android mobile application. For training the image dataset, we make use of MobileNet algorithm that have satisfied us with good result, on top of a Machine Learning Framework, TensorFlow. The percentage of success has been influenced by the image reference size and the optimizer algorithm. The highest performance of implemented model reached 95.13% on its accuracy rate from 23 hand gestures from 13.802 images as dataset, and achieved 100% success on some hand gestures.