Robust PCA-Based Dimensionality Reduction in Human Hand Coordination

Cahree Myrick, Ramana Vinjamuri, Rajasekhar Anguluri

Confirmation: I have read and agree with the IEEE BSN 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: PCA, Robust PCA, Kinematic Synergies, Robust LASSO, Hand Coordination

TL;DR: We show that RPCA combined with RLASSO effectively extracts synergies and reconstructs hand movements even for data containing outliers, outperforming standard PCA and LASSO methods.

Abstract: We consider robust principal component analysis (RPCA) to perform dimensionality reduction for human hand motor control based on kinematic synergies. RPCA decomposes joint angular velocity data into (i) a low-rank matrix capturing coordinated motion patterns and (ii) a sparse matrix isolating sensor artifacts. Next we apply a robust LASSO method to find synergy recruitment weights, yielding a sparse representation of hand grasping tasks. Experiments on 100 grasp trials show that RPCA maintains stable performance up to approximately 20% corruption, while classical PCA degrades quickly. Our results suggest that robustifying standard PCA and LASSO enables reliable synergy extraction even with inexpensive, low-quality sensors, supporting affordable experimentation and improved prosthetic or robotic hand control.

Track: 3. Signal processing, machine learning, deep learning, and decision-support algorithms for digital and computational health

NominateReviewer: Rajasekhar Anguluri and [e-mail]

Submission Number: 127