Deteksi Bahasa Isyarat Berdasarkan Abjad Menggunakan Metode LSTM (Long Short Term Memory)

The LSTM based sign language detection system combined with the use of mediapipe can recognize hand gestures in real time with high accuracy. Alphabet based sign language can use this model to collect temporal patterns of hand gestures. The data used in this study are 30 sample for each alphabet based on American Sign Language (ASL). The data is processed through landmark detection on the hand using mediapipe and opencv, keypoints extraction, folder creation and pre – processing with 80% data divisior for training data and 20 % data for testing data. Using Adam optimization, categorical crossentropy loss function and evaluation metric. The model was trained with 100 epochs and evaluated using confusion matrix. The evaluation results showed that the LSTM model performed well with an accuracy of 98%.