Localization (determination of the position of each sensor node) of a large scale sensor network is an important issue for applying network sensing technique to the health monitoring of the infrastructures. This paper presents an implementation of acoustic ranging and proposes a distributed algorithm for localization of sensor network. Relative positions between sensor nodes are estimated based on acoustic ranging through the inverse Delaunay algorithm. This algorithm localizes all the nodes simultaneously, thus, the accumulation of the error in the localization is suppressed. The robustness and the scalability of this algorithm are examined through numerical simulations. Noise tolerant acoustic ranging algorithm that employs digital signal processing techniques is implemented in an off-the-shelf sensor platform (Mica2). Experiments show that this acoustic ranging algorithm is sufficient to give average range estimation error below 10 cm. Together with this acoustic ranging algorithm, the inverse Delaunay algorithm is also implemented in Mica2. This enables automatic localization of sensor network. Field experiment was conducted to evaluate the accuracy of the localization of the system.