In many applications based on Wireless Sensor Networks (WSNs) with static sensor nodes, the availability of accurate location information of the network nodes may become essential. The node localization problem is to estimate all the unknown node positions, based on noisy pairwise distance measurements of nodes within range of each other. Maximum Likelihood (ML) estimation results in a non-convex problem, which is further complicated by the fact that sufficient conditions for the solution to be unique are not easily identified, especially when dealing with sparse networks. Thereby, different node configurations can provide equally good fitness results, with only one of them corresponding to the real network geometry. This paper presents a novel soft-computing localization technique based on hybridizing a Harmony Search (HS) algorithm with a local search procedure whose aim is to identify the localizability issues and mitigate its effects during the iterative process. Moreover, certain connectivity-based geometrical constraints are exploited to further reduce the areas where each sensor node can be located. Simulation results show that our approach outperforms a previously proposed meta-heuristic localization scheme based on the Simulated Annealing (SA) algorithm, in terms of both localization error and computational cost.