LLM-Conditioned Synthesis of Pathological Gaits via Structured Gait-Language Representations
For researchers in gait analysis and rehabilitation, this provides a data augmentation method to address data scarcity, though the improvement is incremental.
Pathological gait datasets are scarce due to privacy and cost. The authors propose an LLM-guided framework to synthesize 3D gait sequences from text, achieving 92.77% accuracy on a GRU classifier trained with real and synthetic data under leave-one-subject-out protocol.
Pathological gait datasets remain scarce due to privacy, recruitment, cost, and movement variability. Our work presents a multimodal LLM-guided framework for pathology-aware 3D gait data synthesis from structured textual descriptions. The proposed method generates fixed-length synthetic skeleton-based gait sequences for pathological gait classification tasks. The framework combines motion tokenisation, pathology-aware language conditioning, LLM-based semantic augmentation, and language-to-gait generation. A key contribution is the proposed pathological tokeniser, which is designed to preserve pathology-specific motion characteristics during discrete representation learning. Experiments suggest that the proposed synthetic sequences improve downstream classification for recurrent classifiers when combined with real data. The best result is obtained using a GRU classifier trained with real and synthetic samples, achieving 92.77\% accuracy under a leave-one-subject-out protocol.