Speaker
Description
Speech directivity has been methodically studied in theatrical design for centuries, and more precisely characterised using electroacoustic measurements for nearly a century. More recently, measurement systems comprising several dozen microphones have enabled detailed characterization of individual phonemes across different languages. It nonetheless remains unclear how the directivity of individual phonemes relates to that of running speech. To investigate such differences, 26 participants produced 9 vowels, 12 sustained consonants, and 6 phonetically balanced sentences in French. Measurements were carried out using a circular array of 180 MEMS microphones at an angular resolution of 2°, enabling high spatial resolution investigation of directivity patterns in the horizontal plane. Resulting directivity patterns are compared as a function of frequency band in terms of inter-channel level differences, as well as through circular harmonics decomposition. This preliminary work will be followed by subsequent studies comparing with other languages. In the context of immersive audio simulations, results of this study can be used for perceptual studies examining if a single general directivity pattern for running speech is sufficient, or whether phoneme-dependent directivity variations would improve rendering quality.