|
Rehabilitation aims to assist individuals in recovering or enhancing functions that have been lost or impaired due to injury, illness, or disease. The automatic assessment of physical rehabilitation exercises offers a valuable method for patient supervision, which can complement or potentially substitute traditional clinical evaluations. On the other hand, the challenges associated with acquiring large-scale annotated datasets instigate the need for self-supervised learning and transfer learning in the rehabilitation domain. Our proposed approach integrates these two strategies through Low-Rank Adaptation (LoRA) for both pretraining and fine-tuning. Specifically, we train a foundation model to learn robust 3D skeleton features tailored to varying levels of masked motion complexity. In the first stage, we train a transformer-based model with a graph embedding layer. As training progresses, we decrease the ratio of masked joints and introduce additional layers to the model architecture at each stage. By incorporating LoRA layers, we enable the extraction of distinct features for each masking level without significantly increasing the model size. Fine-tuning for downstream tasks demonstrates that the model performs better when different masked motion levels are utilized. Through extensive experiments conducted on the publicly available KIMORE and UI-PRMD datasets, we demonstrate the effectiveness of our approach in accurately evaluating the execution quality of rehabilitation exercises, surpassing state-of-the-art performance across all metrics. Additionally, we validate the effectiveness of our self-supervised learning approach in action recognition tasks, achieving high accuracy on NTU-60 dataset with a compact model.
|
