Spectral differences between live and replayed speech measured in occluded ear.

Abstract

To further our understanding of the transfer function between the mouth and the eardrum for occluded ears, we compared the recordings of live and played-back self-generated speech captured in the speaker’s ear canals. The airborne speech feedback (during the speech production) and its recording (during the playback phase) were delivered binaurally via earmolds or insert earphones (ER-2, Etymotic Research, Inc.). Three earmold styles were tested: Full shell with the canal portion extending either to the first or the second bend of the participant’s ear canal, and a custom open-fit (Microsonic 19A). In four separate conditions, in-the-canal speech recordings were made using a pair of probe tube microphones (ER-7C, Etymotic Research, Inc.) that were inserted at a fixed distance from the eardrum. Two female speakers provided samples of four word-imbedded vowels (heed, head, hawed, who’d). The participants spoke at a comfortable speaking level, and their speech was presented at a sound pressure level calibrated to match the level at the ears. The equivalent sound level and long-term average spectra of the corresponding pairs of speech samples were then analyzed as a function of the ear occlusion type. [Support provided by the Rehabilitation Engineering Research Center (RERC) on Hearing Enhancement.]