Near-ear sound pressure level distribution in everyday life considering the user’s own voice and privacy

Recently, exploring acoustic conditions of people in their everyday environments has drawn a lot attention. One the most important and disturbing sound sources is test participant’s own voice. This contribution proposes an algorithm to determine own-voice audio segments (OVS) for blocks 125 ms method measuring pressure levels (SPL) without violating privacy laws. The voice detection (OVD) here developed based on machine learning set features that do not allow speech reconstruction. A manually labeled real-world recording one full day showed reliable robust results. Moreover, OVD was applied 13 near-ear recordings hearing-impaired participants ecological momentary assessment (EMA) study. analysis shows grand mean percentage predicted OVS during approx. 10% which corresponds well other published data. These had small impact median SPL over all However, short intervals, significant differences up 30 dB occurred measured SPL, depending proportion background noise.