Modeling of physiological tremor with quaternion variant of extreme learning machines

Abstract

Hand-held robotic surgical instruments are developed to acquire the maneuvered hand motion of the surgeon and then provide a control signal for real-time compensation of the physiological tremor in three-dimensional (3-D) space. For active tremor compensation, accurate modeling and estimation of physiological tremor is essential. The current modeling techniques that models tremor in 3D space consider the motion in three-axes [x, y, and z axes) as three separate one-dimensional signals and then perform modeling separately. Recently, it has been shown that for physiological tremor motion there exists cross dimensional coupling and it improves the modeling accuracy. Motivated by this, a quaternion variant for extreme learning machines is developed for accurate 3D modeling of tremor. The developed method is validated with real tremor data and the obtained results highlighted the suitability of this method for accurate tremor modeling in 3D space.