Classification using Efficient LU Decomposition in Sensornets

We consider the popular application of detection, classification and tracking and their feasibility in resource constrained sensornets. We concentrate on the classification aspect, by decomposing the complex, computationally intensive signal processing Maximum-APosterior (MAP) classifier into simpler computationally and communicationally load balanced procedures, using a clustering approach. LU decomposition is an efficient approach for computing the inverse of covariance matrices required in the MAP classifier. We thus explore feasibility of LU decomposition in sensornets. We present power-aware and load balanced techniques for LU decomposition of the covariance matrices in sensornets alongwith their analytical and power consumption analyses.