Coding theory for reliable signal processing

With increased dependence on technology in daily life, there is a need to ensure their reliable performance. There are many applications where we carry out inference tasks assisted by signal processing systems. A typical system performing an inference task can fail due to multiple reasons: presence of a component with permanent failure, a malicious component providing corrupt information, or there might simply be an unreliable component which randomly provides faulty data. Therefore, it is important to design systems which perform reliably even in the presence of such unreliable components. Coding theory based techniques provide a possible solution to this problem. In this position paper, we survey some of our recent work on the use of coding theory based techniques for the design of some signal processing applications. As examples, we consider distributed classification and target localization in wireless sensor networks. We also consider the more recent paradigm of crowdsourcing and discuss how coding based techniques can be used to mitigate the effect of unreliable crowd workers in the system.