Speech privacy through dynamic fundamental frequency matching

The lack of speech privacy is known to be detrimental to one's physical and psychological health. When physical barriers are not feasible to provide this, a masking noise may be implemented to provide speech privacy. Such maskers are most effective when the difference in mean fundamental frequency (F0) between speech and the masker is minimised. Furthermore, noise annoyance is primarily attributed to the loudness and presence of high-pitched components. The present study investigates the effectiveness and annoyance caused by a masker that minimises instantaneous F0 differences. Two seed maskers (pink and babble noise) were used to generate a unique masker for each test sentence read by a variety of speakers. A subjective listening test and annoyance survey were carried out to determine the effects of the dynamically matched masker, a stationary masker at the mean fundamental frequency of each speaker, and the seed masker without F0 processing. The maskers were played at a variety of target to masker ratios. Preliminary results show that the dynamically matched masker is as effective and less annoying than the stationary masker.