New techniques for measuring speech privacy and efficiency of sound masking systems

Abstract

Traditionally, speech privacy has been measured by means of the Articulation Index (AI), transposed to rate privacy rather than intelligibility (PI=1-AI). However, this is an indirect and slightly cumbersome method that usually requires a spreadsheet calculation to yield the Privacy Index rating. Furthermore, the traditional AI approach does not take into account room reverberation effects (though the very much more complex SII does). The paper discusses the potential use of STI and STIPa as direct measures of speech privacy and shows that, potentially, they may be able to offer a superior approach. The benefits and limitations of the methods are highlighted together with the results from a number of case studies. It is concluded that while the method has potential merit, a number of the limiting factors require further research. It particular, it is shown that the index scale at lower levels of intelligibility may need to be revised.