Abstract: In order to achieve fine operation of dextrous robotic hand, the features of tactile data time-series of dextrous hand are analyzed, and a bag of models based on piecewise linear dynamic systems is proposed. Furthermore, corresponding relation between tactile data time-series and family of models describing different types of objects is established in low-dimensional feature space via solving parametric matrixes of dynamic system models. Some experiments are conducted on a dextrous hand with tactile array sensor, and the results show that the proposed method can give not only a precise description of the dynamic process corresponding to a same type of objects, but also a comprehensive description corresponding to different types of objects, which will lay the foundation for multi-modal information fusion in feature space.