Research Of Filtering Algorithm Based On Support Vector Machine

Along with the development of medical techniques, minimally invasive surgery is becoming more and more prevalent, and being widely used in clinics. Nowadays, the robotic system is introduced in MIS, the stability, accuracy and controllability of microsurgery are greatly improved. Microsurgery robot has been research focus in medical robot field and many outstanding attainments have been achieved in this field.Considering surgeons always contribute to the microsurgery robotic system, the stability, accuracy of minimally invasive surgery may be influenced by hand tremor from surgeons. Therefore, in order to ensure the quality of microsurgery, based on support vector machine, a novel tremor filter is proposed for tremor in this paper.1. The kinematic model and tremor suppression strategies of master-slave robot were discussed in this paper. This paper studies the advantages and disadvantages among neural network, WFLC, BMFLC and SVM.2. Aiming at the limitation of number of sample for microsurgery robot, support vector machine which is based on small samples and higher space is introduced. Also, this paper analyzes the mathematic model and physical model of tremor filter. Considering the fact that hand tremors have different frequencies and amplitudes, different learning weights are designed for kinds of hand tremors. A linear learning-weight function based on time sequence is designed for the tremor with low frequency, and a Gaussian learning-weight function is devised for tremor with high frequency. Based on the SVM theory, TSF-SVM can solve the problem of nonlinear system sampling with small scale training data or high dimensional input space.3. Considering without the parameters guidance of unified theory in TSF-SVM and in order to enhance the security and accuracy of tremor filter and reduce the disturbance of one’s experience, particle swarm optimization was introduced in this paper. Through presentation of particle, particle update and fitness function, it is employed to finding the optical parameters combination of tremor filter.4. Simulation experiments using various kinds of tremor signal for the proposed models are given in this paper. The experiment results demonstrate that the proposed tremor filters have good results for suppressing tremor signal.