Near-Field Evaluation of Reproducible Speech Sources

Abstract

The spatial speech reproduction capabilities of a KEMAR mouth simulator, a loudspeaker, the piston on sphere model and a circular harmonic fitting are evaluated in the near-field. The speech directivity of 24 human subjects, both male and female, is measured using a semicircular microphone array of radius 36.5 cm in the horizontal plane. Impulse responses are captured for the two devices and filters are generated for the two numerical models to emulate their directional effect on speech reproduction. The four repeatable speech sources are evaluated through comparison to the recorded human speech both objectively, through directivity pattern and spectral magnitude differences, and subjectively, through a listening test on perceived coloration. Results show that the repeatable sources perform relatively well under the metric of directivity but irregularities in their directivity patterns introduce audible coloration for off-axis directions.