By way of tests, we studied the possibility of estimating speech intelligibility in noisy rooms (factory-type rooms). The results are patterned by a regression function of the variable "Signal-to-Noise Ratio". It is supposed that the model depends on the physical variables involved in the process such as the diffusers (by the mean of spectrum and directivity) or the nature of the rooms. First, we plan to characterize the intelligibility through these parameters and the Signal-to-Noise ratio, our goal being to optimize the scores at each place in the room. Following some recent studies, we take into account that the first x milliseconds of the impulse response are "useful" for intelligibility. This leads to the definition of an equivalent Signal-to-Noise ratio. Doing so, the influence of the room by its impulse response is included in our results. The study of the model will be reduced to the analysis of only one parameter binded with the diffuser. The results show the correlation between this parameter and the directivity coefficient.
Cite as: Randrianarison, A.L., Legros, C. (1992) Modeling speech intelligibility under reverberation and noise: room and diffuser influences. Proc. ETRW on Speech Processing in Adverse Conditions, 65-68
@inproceedings{randrianarison92_spac,
author={A. L. Randrianarison and C. Legros},
title={{Modeling speech intelligibility under reverberation and noise: room and diffuser influences}},
year=1992,
booktitle={Proc. ETRW on Speech Processing in Adverse Conditions},
pages={65--68}
}