Malaysian Sign Language Alphabet Recognition Through Hand Landmarks Feature and Machine Learning

The population of hearing and speaking impairment often experiences difficulties when others fail to understand them and vice versa. This presents a major issue in terms of humanitarian concerns and requires a solution that is specific to the local context. Although some solutions to this challenge have been developed through the use of wearable sensors and convolution neural networks, the programming procedure for this technology is quite complicated. Therefore, to solve this problem this study implements a different approach by using artificial neural network (ANN) based machine learning model to identify Malaysian Sign Language (MSL) from coordinates difference between datapoints and the offset datapoint on the hand landmark model. To assess the model’s efficiency, several metrics such as accuracy, precision, recall, f1-score, support, and confusion matrix are recorded. The results showed that the ANN model performed well in identifying the MSL with accuracy rates of more than 90% reached with just minor fluctuations.