Double entendre: Embedding a secondary message in pointillistic speech

Abstract

A dual-purpose application for pointillistic speech (Kidd et al., 2009, JASA 126, EL196-201) was described. speech was represented by a time-frequency matrix of pure-tone derived from the original stimulus. pattern of frequencies and intensities of the points coded the primary message while the phases of the points coded an independent secondary message. For example, the audible and intelligible word shoes may also convey the (unintelligible) ASCII characters of the phrase The road goes ever on and on… by the phase-coded bit pattern of the pointillistic representation of shoes. success in recovering the secondary message via signal processing was examined while the intelligibility of the primary message in speech recognition experiments was measured. Tradeoffs for accuracy in primary and secondary message transfer/reception were explored for both pointillistic speech and hybrid speech comprising natural and pointillistic representations. Some initial observations about the possible uses and limitations of this approach were considered. Near-perfect accuracy for both message types was found for several hybrid combinations supporting the general feasibility of this method.