Body Sensor Networking, Design and Algorithms

10Signal Processing for Sensor Networks

10.1 Introduction

In the introductory chapters of this book it has been emphasised and clarified that there are three major components of sensing, processing, and communications involved in the design of sensor networks. Dealing with conventional computational machines, electronic gadgets, and pervasive systems, the sensor data need to be digitised before applying any learning or processing to them. Analogue-to-digital convertors are therefore embedded within all the data acquisition cards as well as sensor motes.

Although many hardware and software platforms have been introduced (some are addressed in Chapter 14 of this book) for hosting the necessary processing algorithms, yet the fundamentals, advances, and the applications of effective algorithms for sensor networks have not been explored in detail. Such algorithms have roots in fundamental digital signal processing concepts for sensor network theories.

Regarding the techniques in machine learning, most of the existing algorithms rely on a central system for fusing and aggregating the collected information. The concepts have been further advanced by introducing the decentralised systems and the notion of cooperative networks very recently. This requires migration from fusion, which is necessary for centralised systems, to consensus and diffusion techniques, deployed in decentralised networks. In decentralised systems, the sensors are smarter and act as intelligent agents capable of ...