Physiological Tremor Modeling with Singular Spectrum Analysis-Based Quaternion Extreme Learning Machine for Handheld Surgical Robotics

Hand-held robotic surgical instruments are designed to capture the surgeon’s hand movements and generate control signals for the real-time compensation of physiological tremors in three-dimensional (3D) space. Accurate modeling and estimation of physiological tremor are essential for effective active tremor compensation. Existing techniques for 3D tip position control typically model and cancel tremors independently along the x, y, and z axes, thereby neglecting the dynamic coupling among these three dimensions. We hypothesized that a system incorporating this coupling information could model tremor more accurately than existing methods. Based on this, we propose a novel approach that integrates singular spectrum analysis with a quaternion extreme learning machine (SSA-QELM) to accurately estimate voluntary and tremor motion. The proposed SSA-QELM was validated using real tremor data, and the results demonstrated its effectiveness in accurately modeling tremors in 3D space.