Filtragem de Ruído Semântico em Dados Cinemáticos Pós-AVC: Uma Abordagem Automatizada de Aprendizado Profundo

• Victor Y. K. Kwon UFES
• Leonardo C. D. Rocha UFSJ
• Diego R. C. Dias UFES

DOI: https://doi.org/10.5753/webmedia_estendido.2026.33370

Resumo

Post-stroke rehabilitation increasingly relies on motion capture and extended reality (XR) technologies. However, datasets collected in these environments, such as those stored in the Rehabilitation Database (ReBase), frequently suffer from noise contamination. A critical challenge in this domain is distinguishing clinically relevant variations, such as compensatory movements, from spurious artifacts caused by gestures unrelated to the therapeutic task. This work proposes an automated and computationally efficient filtering approach based on a multilayer perceptron (MLP). Unlike complex recurrent architectures, the proposed MLP captures gait patterns with observation intervals of just 5 seconds, making the solution computationally accessible and discriminative according to the adopted criteria. The model was trained with knee angle time series labeled by heuristic thresholds, achieving an accuracy of 93.18% and a precision of 100% (zero false positives) on the evaluated test set. It is noteworthy that the absence of false positives is restricted to the analyzed partition and does not guarantee the total elimination of artifacts in future data, nor does it constitute independent clinical validation. Evaluations on unseen recordings offer qualitative support for the model’s ability to filter data without manual per-patient calibration. Replacing manual curation with this automated screening enhances ReBase’s scalability, increases storage efficiency, and supports the monitoring of motor recovery protocols.

Palavras-chave: ReBase, Neural Networks, Data Filtering, Neurological Rehabilitation, Body Kinematics

Referências

Ferdous Ahmed et al. 2019. Kalman filter-based noise reduction framework for posture estimation using depth sensor. In 2019 IEEE 18th International Conference on Cognitive Informatics & Cognitive Computing (ICCI* CC). IEEE, 150–158.

Luiz Felipe de Camargo, Gabriel Rissatti Miguel, Juliana da Costa Feitosa, Luis Guilherme Silva Rodrigues, Diego Roberto Colombo Dias, and José Remo Ferreira Brega. 2025. Combining Computer Vision and Artificial Intelligence to Support the Rehabilitation of Post-stroke Patients. In International Conference on Computational Science and Its Applications. Springer, 415–432.

Marianna Capecci et al. 2019. The KIMORE dataset: Kinematic assessment of movement and clinical scores for remote monitoring of physical rehabilitation. IEEE Transactions on Neural Systems and Rehabilitation Engineering 27, 7 (2019), 1436–1448.

Diego RC Dias, Iago C Alvarenga, Marcelo P Guimarães, Luis C Trevelin, Gabriela Castellano, and Alexandre F Brandão. 2018. eStreet: virtual reality and wearable devices applied to rehabilitation. In International Conference on Computational Science and Its Applications. Springer, 775–789.

Yalin Liao, Aleksandar Vakanski, and Min Xian. 2020. A deep learning framework for assessing physical rehabilitation exercises. IEEE Transactions on Neural Systems and Rehabilitation Engineering 28, 2 (2020), 468–477.

Jun Liu et al. 2019. NTU RGB+ D 120: A large-scale benchmark for 3D human activity understanding. IEEE Transactions on Pattern Analysis and Machine Intelligence 42, 10 (2019), 2684–2701.

M. Maier et al. 2019. Effect of specific over nonspecific VR-based rehabilitation on poststroke motor recovery: a systematic meta-analysis. Neurorehabilitation and Neural Repair 33, 2 (2019), 112–129.

Alina Miron et al. 2021. IntelliRehabDS (IRDS) – A dataset of physical rehabilitation movements. Data 6, 5 (2021), 46.

Elvis Ribeiro, Alexandre Brandão, Marcelo Guimarães, Leonardo Rocha, José Remo Brega, Rodolfo Villaça, and Diego Dias. 2025. A Body Tracking Framework for Neuromotor Rehabilitation with Multimedia Application Support. In International Conference on Computational Science and Its Applications. Springer, 345–360.

L. G. S. Rodrigues et al. 2022. A lower limb mixed reality system for the elderly. In Proceedings of the 24th Symposium on Virtual and Augmented Reality. ACM, New York, 5. DOI: 10.1145/3604479.3604531

Mariano Tepper and Guillermo Sapiro. 2013. Fast L1 smoothing splines with an application to Kinect depth data. In 2013 IEEE International Conference on Image Processing. IEEE, 504–508.

T. Trotta et al. 2024. ReBase: a framework for acquiring and managing body tracking data. In Proceedings of the Symposium on Virtual and Augmented Reality. ACM, New York, 9. DOI: 10.1145/3691573.3691578

Aleksandar Vakanski, Hyung-pil Jun, David Paul, and Russell Baker. 2018. A data set of human body movements for physical rehabilitation exercises. Data 3, 1 (2018), 2.

David A Winter. 2009. Biomechanics and motor control of human movement. John Wiley & Sons.

Sijie Yan, Yuanjun Xiong, and Dahua Lin. 2018. Spatial temporal graph convolutional networks for skeleton-based action recognition. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 32.