In wireless sensor network (WSN), location information is required to achieve the goal in many applications. Most of existing localization algorithms for mobile sensor networks are based on the Sequential Monte Carlo (SMC) method. SMC-based methods has either high localization error in low-anchor density environments or high communication cost. In this paper, we propose a non-SMC-based distributed localization scheme. Our main idea comes from the observation that a sensor node which hears another node for a sufficiently long time has trajectory similar to the heard node's trajectory. Simulation shows that our scheme has low localization error, low communication cost, and low computation cost.