Video-Based Sign Language Recognition Without Temporal Segmentation