In several wireless sensor network applications the availability of accurate nodes' location information is essential to make collected data meaningful. In this context, estimating the positions of all unknown-located nodes of the network based on noisy distance-related measurements (usually referred to as localization) generally embodies a non-convex optimization problem, which is further exacerbated by the fact that the network may not be uniquely localizable, especially when its connectivity degree is not sufficiently high. In order to efficiently tackle this problem, we propose a novel two-objective localization approach based on the combination of the harmony search (HS) algorithm and a local search procedure. Moreover, some connectivity-based geometrical constraints are defined and exploited to limit the areas in which sensor nodes can be located. The proposed method is tested with different network configurations and compared, in terms of normalized localization error and three multi-objective quality indicators, with a state-of-the-art metaheuristic localization scheme based on the Pareto archived evolution strategy (PAES). The results show that the proposed approach achieves considerable accuracies and, in the majority of the scenarios, outperforms PAES.