Fountain Data Estimation within Bayesian model Classification in Wireless Sensor Network

Abstract

In this paper, a novel distributed estimation scheme is proposed. This model combines learning methods and fountain codes. In order to minimize the number of transmissions as well as the impact of useless data, we determine the optimal minimal number of encoded packets needed for a successful decoding. Sensor observations are encoded using fountain codes. Then messages are collected at the cluster head where a final estimation is provided with a classification based on Bayes rules. The main goal of this paper is to estimate the needed number of encoded packets according to a Bayesian method. The performance results have been analyzed through a comparison with the Support Vector Machine.