Wavelet Kernel Based on Identification for Nonlinear Hybrid Systems

Abstract

This paper presents a new method based on wavelet for a class of nonlinear hybrid systems identification. Hybrid systems identification is composed of two problems; estimate the discrete modes or switch among the system modes and estimate continues submodels. In this paper, we assumed that haven’t any prior knowledge about data classification and submodels identification. Also the combining of feature vector selection algorithm and wavelet are used in subspace learning and support vector machine as a classifier. The results indicate that the error of using the wavelet in subspace learning process becomes low. In addition, the proposed method is convergent and has an acceptable response in presence of high-power noise.