SignDEG (DINO Eats Geometry) is a temporal sign language encoder that incorporates articulated hand geometry as a pre‑training target while using only video at inference. It leverages frozen DINOv3 features of tracked hands and faces, feeding them into per‑stream temporal Transformers, and introduces two geometric objectives—Geometric Latent Prediction (GLP) and Geometric Relation Transfer (GRT)—to supervise intermediate and final blocks. Jointly trained with temporal self‑distillation, SignDEG achieves higher BLEU‑4 scores on How2Sign/OpenASL and improves recall on ASL Citizen and Sem‑Lex datasets compared to SHuBERT, demonstrating that articulated structure can be transferred into an appearance encoder without requiring a hand reconstruction model at deployment.
arXiv:2609.14122v1 Announce Type: new Abstract: We study the challenge of sign language video mimicking: given a driving video and a single reference frame, synthesize a video where the target signer...
SignMatch introduces a prototype‑structured embedding space that learns to match dictionary sign videos with continuous sign language footage based solely on visual similarity of handshape and motion. By mapping isolated dictionary exemplars into this space, the method enables direct, embedding‑based sign matching and can generalise to unseen signs using only dictionary examples. Experiments on ASL‑Citizen, ChaLearn OSLWL, and BOBSL CSLR2 benchmarks show strong cross‑dataset, cross‑task, and cross‑language performance, outperforming prior approaches on American, British, and Spanish sign languages without benchmark‑specific supervision.
arXiv:2605. 01720v3 Announce Type: replace-cross Abstract: Existing large-scale sign language resources typically provide supervision only at the level of raw video-text alignment and are often produced in laboratory settings.
arXiv:2608. 06407v1 Announce Type: cross Abstract: Automated Sign Language Recognition for under-represented languages remains a largely unsolved problem.
SignRR is a new sign language production framework that combines retrieval of real sign motion segments with a learned refinement step to produce globally coherent signing sequences. It starts from a dictionary of authentic sign segments and refines them using a part-aware Residual VQ‑VAE, preserving fine hand articulation while handling temporal length differences in latent space. Experiments on PHOENIX14T and CSL‑Daily demonstrate state‑of‑the‑art back‑translation performance and competitive pose quality.
arXiv:2609.12993v1 Announce Type: cross Abstract: We investigate the effect of pretrained T5 model scale and explicit motion features on pose-to-text Indian Sign Language Translation (SLT) for the WS...
M3T introduces a discrete multi‑modal motion token system for sign language production, addressing the need for non‑manual features such as mouthings, eyebrow raises, gaze, and head movements. The approach couples FLAME’s expressive facial space with SMPL‑X body parameters and uses modality‑specific Finite Scalar Quantization VAEs to achieve high face codebook utilization (99.0%). Trained with an autoregressive transformer and a sign‑to‑text translation objective, M3T outperforms existing methods on three standard datasets, notably improving accuracy on NMFs‑CSL from 49.0% to 58.3% without large‑scale pre‑training.
arXiv:2204. 02803v2 Announce Type: replace-cross Abstract: Sign language recognition from monocular video or 2D pose sequences is challenging, both because 3D information must be inferred from 2D observations and because the signal is inherently spatiotemporal.
The paper introduces the Sequential Spatio-Temporal Attention Network (SSTAN), a Transformer-based architecture that replaces traditional Graph Convolutional Networks for sign language recognition. SSTAN uses a hierarchical, stacked design with Spatial Multi-Head Attention to model joint relationships within frames and Temporal Multi-Head Attention to capture long-range dependencies across frames, eliminating the need for predefined skeletal graphs. Experiments on large-scale datasets (WLASL, JSL, KSL) show that SSTAN, trained from scratch, achieves state‑of‑the‑art performance in fingerspelling categories and outperforms other skeleton‑only methods on WLASL, highlighting its data efficiency and ability to learn complex spatio‑temporal patterns.
Accurate monocular 4D hand reconstruction remains challenging. Per-frame discriminative regressors lack temporal context and often produce jittery predictions.
arXiv:2605. 05367v2 Announce Type: replace-cross Abstract: Existing 3D sign language avatar reconstruction methods are developed and evaluated exclusively on Western sign languages, and no 3D parametric annotations exist for any Arabic Sign Language dataset, a gap that blocks the development of avatar-based accessibility applications for the Arab Deaf community.
arXiv:2607. 11221v1 Announce Type: cross Abstract: Accurate monocular 4D hand reconstruction remains challenging.