Data Extraction from Wireless Sensor Networks Using Fountain Codes

We propose a universal and energy efficient method of data extraction from a wireless sensor network (WSN), based on digital fountain codes (DFCs) and joint-source channel decoding. We consider a WSN that aims to recreate a binary field at a distant fusion center (FC). Our method (i) exploits the feedback channel from the FC to the WSN to implement a distributed 'rate-less' DFC that automatically tunes the number of transmissions to the channel capacity, and (ii) treats the spatiotemporal dependencies in the natural field as an outer code, and jointly decodes this concatenation at the FC. For moderate distortion the proposed procedure achieves an energy efficiency close to the rate-distortion limit, while retaining the important property of having a computationally simple encoder invariant to changes in the statistical model of the source or the channel.