Towards a method for loudness-based analysis of the sound of one's own voice

This paper outlines the steps in objectively estimating the time-varying loudness of one's own voice in a room (i.e. autophonic loudness). Voice recordings, made with a near-mouth microphone, are converted to the sound that reaches the two eardrums of the talking (or singing)-listener by convolving them with the impulse responses from the mouth to the respective ears of an anthropomorphic head and torso. The influences of bone-conducted sound and room reflections are taken into account. These convolved recordings are then processed with a computational time-varying loudness model. The method is demonstrated by a short case study, and the results illustrate something of the benefit of loudness analysis over sound pressure level analysis for representing autophonic loudness.