Multiterminal Source Coding With Copula Regression for Wireless Sensor Networks Gathering Diverse Data

TL;DR: A novel multiterminal source code design is proposed, which, contrary to prior work, utilizes both the intra- and the inter-sensor data dependences, and makes use of copula functions, namely, a statistical model that captures the dependence structure amongst heterogeneous data types.

Abstract: Efficient data compression at a low processing and communication cost is a key challenge in wireless sensor networks. In this paper, we propose a novel multiterminal source code design, which, contrary to prior work, utilizes both the intra- and the inter-sensor data dependences. The former is exploited by applying simple differential pulse-code modulation followed by arithmetic entropy coding at each distributed encoder. This approach limits the encoding complexity and provides for a flexible design that adapts to the variations in the number of operating sensors. Moreover, we propose a regression method applied at the joint decoder, which aims at leveraging the inter-sensor data dependences. Unlike existing work that focuses on homogeneous data types, the proposed method makes use of copula functions, namely, a statistical model that captures the dependence structure amongst heterogeneous data types. Experimentation using real sensor measurements—taken from the Intel-Berkeley database—shows that the proposed system achieves significant compression improvements compared with the state-of-the-art multiterminal and distributed source coding schemes.