Scholarlink

<p>An effective rehabilitation program can significantly hasten the recovery of patients. It promotes the metabolism of damaged tissues and aids in the seamless integration of morphology, function, and structure. With the emergence of sophisticated computer vision technologies, rehabilitation training has become more feasible. Depth imaging devices or traditional video sensors can collect movement data, which can then be automatically evaluated to provide quantitative feedback. The movement quality assessment (MQA) problem has been a subject of prior research. This paper presents a novel methodology that includes three innovative skeletal data augmentation techniques and an effective scoring model, named the Residual Movement Metric Network (R2MN). The data generated through these proposed techniques led to a substantial improvement in results. Defining MQA scores is crucial for solving the MQA problem. The proposed scoring model can easily assess the quality of an action and provide relevant metrics. Compared to the existing scoring system, the proposed model shows an 11.4% improvement in prediction results. Finally, a new transformer-based MQA architecture is put forward. In contrast to existing methods, the prediction of movement quality scores on the UI—PRMD dataset is enhanced by 34%, and on the KIMORE, dataset is enhanced by 34.5%.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Residual Movement Metrology Network and Transformer-Based Framework for Evaluating Movement Quality

  • 期刊论文

  • 发表日期：

    2025年4月23日

查看全文
  • WeiQing Meng,
  • Peng Chen,
  • Jun Zhang,
  • Yi Xia,
  • Bing Wang

摘要

An effective rehabilitation program can significantly hasten the recovery of patients. It promotes the metabolism of damaged tissues and aids in the seamless integration of morphology, function, and structure. With the emergence of sophisticated computer vision technologies, rehabilitation training has become more feasible. Depth imaging devices or traditional video sensors can collect movement data, which can then be automatically evaluated to provide quantitative feedback. The movement quality assessment (MQA) problem has been a subject of prior research. This paper presents a novel methodology that includes three innovative skeletal data augmentation techniques and an effective scoring model, named the Residual Movement Metric Network (R2MN). The data generated through these proposed techniques led to a substantial improvement in results. Defining MQA scores is crucial for solving the MQA problem. The proposed scoring model can easily assess the quality of an action and provide relevant metrics. Compared to the existing scoring system, the proposed model shows an 11.4% improvement in prediction results. Finally, a new transformer-based MQA architecture is put forward. In contrast to existing methods, the prediction of movement quality scores on the UI—PRMD dataset is enhanced by 34%, and on the KIMORE, dataset is enhanced by 34.5%.