Using a Discriminator to Improve Compressive Sensing Efficiency

Abstract

Our work defines, implements, and evaluates a modification to a spectrum-based compression scheme for data streams coming from jet aircraft health-monitoring sensors. The modification consists of the addition of a discriminator which separates data streams into similar classes. We create and justify a simulation of a jet sensor network as a source for data streams. The data streams are compressed and decompressed under the new compression scheme and also under two old ones, and the reconstructions are evaluated for quality. The discriminator-based modification to the existing compression algorithm is found to yield better quality than the other two compression algorithms, at the cost of increased runtime.