A feedforward method based on support vector machine

Abstract

The optimal estimation of the target movement state is the key technology of target tracking. The single-model predictive filtering algorithm cannot satisfy the requirement of accuracy when the system is tracking a strong maneuver target. Because the estimation algorithm is inaccurate when the target's motion state and model do not match. The multi-model algorithm is proposed to solve this problem and it does effectively improve the estimation accuracy. However, it is hard to apply to the actual engineering due to the large calculation quantity. In this paper, a feedforward method based on SVM is proposed. This method uses SVM to classify the target's motion state and gets the most suitable motion model to the targets current state. This model is utilized to predict and estimate the motion parameters of the target in the way of single model. Therefore, the contradiction between calculation quantity and precision is alleviated. Then, the target motion parameters are fed into the system, and the target is tracked with high precision.