An effective tremor-filtering model in teleoperation: Three-domain Wavelet Least Square Support Vector Machine

Abstract

Proper elimination of hand tremors in robot-assisted minimally invasive surgery has become a key concern due to its significant impacts on control precision and surgical success. In this study, a Three-domain Wavelet Least Square Support Vector Machine with Improved Sparrow Search Algorithm (ISSA-TDWLSSVM) model is proposed to forecast and eliminate tremor signals in teleoperation. The multi-domain analysis layer is proposed for redesigning the structure of the LSSVM. The multi-domain analysis layer is divided into three parts, complex-frequency domain, frequency domain, and time domain. The functions corresponding to three domains are the Time series, Mexican hat wavelet, and Gauss–Laplace functions. The other contribution of this study is that a novel wavelet kernel-function is proposed. Meanwhile, an improved sparrow search algorithm is proposed for optimizing the wavelet kernel-function parameter to improve the forecasting accuracy. To prove the generalization ability of the proposed model, the ISSA-TDWLSSVM and existing models are used in three examples. Compared with the existing models, the result showed that the ISSA-TDWLSSVM model has the highest forecasting accuracy. The wavelet kernel function has a good mapping ability. As compared with the LSSVM used in three-axis teleoperation robot, the Accuracy index of the ISSA-TDWLSSVM model is increased by 34.52%. In a word, an effective model is proposed to eliminate tele-operated tremor signals in this study.