Improved DS acoustic-seismic modality fusion for ground-moving target classification in wireless sensor networks

摘要

An improved DS acoustic-seismic modality fusion framework based on cascaded fuzzy classifier (CFC) is proposed to implement ground-moving target classification tasks locally at sensor nodes in wireless sensor networks (WSN). The CFC consists of three and two component binary fuzzy classifiers (BFCs) in seismic and acoustic signal channel respectively. New basic belief assignment (bba) functions are defined for component binary fuzzy classifiers (BFCs) to give out evidences instead of hard decision labels for each unclassified pattern. Available evidences are then combined into a final node classification report using a modified DS method. M-fold cross-validation experiment results show that this implementation gives significantly better performance than the implementation with a majority-voting fusion and a DS fusion implementation with a linear bba function. Performances on different terrains are also given to validate its robustness.