Abstract:

Studies on speech privacy problems in buildings at Bolt Beranek and Newman begun in the mid 1950s showed convincingly that a person's sense of acoustical privacy was directly related to the intelligibility of the intruding speech over the continuous ambient sound present [Cavanaugh et. al., J. Acoust. Soc. Am. 34, 475--492 (1962)]. Both in laboratory and in real-world settings people feel they have confidential privacy when the articulation index (AI) of the intruding speech is 0.05 or less. Higher values of AI are acceptable for less demanding tasks, i.e., normal privacy. The method quantifies each of the significant variables involved and compares a summary of these metrics with response data for prior case histories both to predict, and, in the case of existing situations, to evaluate speech privacy conditions. R. W. Young reviewed the analysis method and showed that sufficient accuracy was preserved if simple, commonly used A-scale sound levels were used in place of more cumbersome frequency-based metrics [R. W. Young, J. Acoust. Soc. Am. 38, 524--533 (1965)]. Later in the 1960s the method was successfully applied to open plan offices which were becoming more prevalent [Hirtle et. al., J. Acoust. Soc. Am. 46. 91A (1969)]. The lessons learned in applying the speech privacy calculation over the years are discussed.