Comparing testing methodologies of speech privacy class in closed offices

Abstract

Speech privacy is a principal research area in speech communication. The term "speech privacy" is generally interpreted a condition where speech cannot be readily understood by, but may be audible to, an unintended listener. The need to prevent sound from intruding into adjacent spaces in both closed and open-plan settings is a concern in various office buildings. Speech privacy is essentially a function of the signal-to-noise ratio, comprising the noise reduction between source and receiver positions, and the masking effective of background noise. Speech Privacy Class (SPC) is one of the commonly used metrics for speech privacy assessment in closed plan offices in North America. However, there is a reluctance to use SPC between closed rooms due to laborious testing requirements as per ASTM E2638; there is a supposed preference to use either Speech Privacy Potential (SPP) based on simpler NIC testing, or to use Articulation Index (AI) despite the AI being created solely for testing open-plan settings. The ASTM E2638-10 standard SPC testing procedure does not assume a diffuse field in the receiving space but rather evaluates the performance at each potential eavesdropping location on the basis that there may be an intentional listener. Thus, to better apply the SPC to unintended listening and speech privacy in typical commercial spaces, the goal is to compare the current methodology as per ASTM E2638-10 with alternative sampling methods considering the talker location and unintended listening positions outside the source room.