Genetic algorithms (GA) have proved to be a robust, efficient search technique for many problems. As the number of variables involved increases, classical GA will often have difficulty and/or require long computational time in obtaining acceptable results. In this paper, a modified GA strategy is proposed to improve the accuracy and computational time for parameter identification of multiple degree-of-freedom (DOF) structural systems. The strategy includes multiple populations or 'species', a search space reduction procedure and new operators designed to provide a robust and reliable identification. Average absolute error in the estimated stiffness values of 1.4% is achieved for a 20-DOF unknown mass system with 5% noise, and even more importantly the maximum error is reduced to only 3.8%.