Dictionary based reconstruction and classification of randomly sampled sensor network data

In this paper, we propose a method for recovering and classifying WSN data while minimizing the number of samples that need to be acquired, processed, and transmitted. The problem is formulated according to the recently proposed framework of Matrix Completion (MC), which asserts that a low rank matrix can be recovered from a small number of randomly sampled entries. The application of MC in WSN data is motivated by the assumption that sensory data exhibit intra-sensor correlations and that these data can be represented using known examples. We formulate the problem as that of recovering the low rank measurement matrix by encoding the contributions of known examples, the dictionary elements, for reconstructing and classifying the data. Experimental results using artificial data suggest that the proposed scheme is able to accurately reconstruct and classify the sensory data from a small number of measurements.