Methods to experimentally characterize the own-voice-generated objective occlusion effect induced by hearables

In this study, the problem of experimentally identifying the own-voice generated objective occlusion effect in hearables is addressed. Challenges arise from the sub-optimal properties of one’s own voice as a test signal, namely, poor reproducibility, limited bandwidth, and the induction of time-variant behavior of the effect being measured. Based on experiments with 19 participants wearing a vented hearable and producing running speech and a sung vowel, it was found that (a) running speech is better suited than vowels in most respects, except for the time-variance of the occlusion effect, (b) the use of transfer function-based estimates of the occlusion effect results in more problems than advantages in comparison to estimates based on power spectral densities, and (c) the popular method of measuring the occlusion effect by simultaneously measuring inside and outside the occluding device entails systematic errors of up to about 3–4 dB, even in the frequency range in which it was previously considered valid. In contrast, the simultaneous measurement with reference to the open contralateral ear is accurate throughout the frequency range in which an acceptable SNR is achieved.