Abstract
The support vector machine (SVM) is proposed for dynamic error correction of measuring systems. The SVM is established based on the structural risk minimization principle rather than minimize the empirical error commonly implemented in the neural networks. Hence, the SVM can overcome the shortcoming of neural networks in dynamic error correction of measuring systems. The feasibility and efficacy of the method are demonstrated by applying it to an example. The results show that the proposed method is still effective even if there is additive measuring noise.