This paper presents a novel approach, based on the theory of hyperplanes, for mode identification of linear systems. The proposed approach can operate on either a set of ordinary differential equations (converted to diagonal form, if needed) or a set of partial fractions derived from a synthesized transfer function of the system under analysis. For either format, the linear system is structured to have as unknown variable a vector containing the residues. Singular value decomposition is initially used to identify an initial sparsity of the residue vector where the number of nonzero values corresponds to the pre-defined order of the dominant poles (eigenvalues) under search. An algorithm based on geometrical search of hyperplanes is used to optimize the selection of the nonzero residue locations, minimizing the residual of the zero residue hyperplanes. Finally, a recalculation of the residues is carried out by using the obtained optimal sparsity.