Robust Keypoint Normalization Method for Korean Sign Language Translation using Transformer

In this paper, We propose a robust keypoint normalization method for sign language translation framework using Transformer. The proposed method normalizes human keypoints based on the length of the neck-shoulder bone. The normalized keypoints are utilized for the Transformer-based framework as an input sequence. In our extensive experiments, the proposed normalization method greatly improves the accuracy and robustness of sign language translation.