An Investigation of Coding Schemes for Sample Entropy Analysis of Communications Data

Abstract

An investigation into the effect of coding scheme choice for sample entropy analysis of communications data is presented. Semantic content and speaker encoding of transmissions are explored. The results suggest a scheme with a greater number of categories will more likely pass surrogate testing, whether through finer distinctions of semantic categorization or through the additions of a speaker-content interaction. The results also suggest the way semantic categories are formed affects how well the scheme permits discrimination between workload conditions with a known effect.