In video-based 2D human pose estimation, existing Transformer models mainly focus on image features, neglecting the implicit information in skeletal connections. This oversight leads to inaccuracies ...