This paper presents a novel approach for the extraction of interpretable rules from piecewise-linear approximation of any nonlinear classifier, e.g. Support Vector Machine (SVM). Recent major interest in kernel methods has drawn an attention to the extraction of human-tractable and interpretable rules from such state-of-art and hyperplane-based classifiers like SVM. Nevertheless, there is no clear understanding of the rule extraction routine from nonlinear SVM. The most comprehensible methods imply very complicated and indirect constraints that strictly enforce convexity of the searched-through half-space of inducted nonlinear classifier [1]. Apparently non-convex hyper-surfaces couldnt be effectively described by a finite set of convex rules without violating bounds of constrained non-convex area. In this paper we describe an approach for converting a widely acknowledged rule extraction algorithm for Linear Support Vector Machine [3] into a number of constrained linear programming (LP) problems. Stated LP problems incorporate finite number of linear constraints based on different individual decision hyperplanes to ensure piecewise-linear approximation of underlying nonlinearity. We claim that proposed approach helps to extract better rules from linearly non-separable cases and could be effectively employed even for non-homogeneous target classes with high inner variance.