In this paper, we propose a quantizer design algorithm that is optimized for source localization in sensor networks. For this application, the goal is to minimize the amount of information that the sensor nodes have to exchange in order to achieve a certain source localization accuracy. We show that this goal can be achieved more efficiently when "application-specific" quantizers are used. Our proposed quantizer design algorithm uses a cost function that takes into account the distance between the actual source position and the position estimated based on quantized data. We also propose a distributed encoding algorithm that is applied after quantization and achieves rate savings by merging quantization bins without any degradation of localization performance. The merging technique in the encoding algorithm exploits the fact that certain combinations of quantization bins at each node cannot occur because the corresponding spatial regions have an empty intersection. We apply these algorithms to a system where an acoustic sensor model is employed for localization. For this case, we introduce the Equally Distance-divided Quantizer (EDQ), designed so that quantizer partitions correspond to a uniform partitioning in terms of distance. Our simulations show the improved performance of our quantizer over traditional quantizer designs. In addition, they show rate savings (32.8%, 5 nodes, 4 bits per node) when our novel bin-merging algorithms are used. Our results also show that an optimized bit allocation leads to significant improvements in localization performance with respect to a bit allocation that uses the same number of bits for each node.
展开▼
