Predictions of Speech Intelligibility with a Model of the Normal and Impaired Auditory-periphery

A fall-off in speech intelligibility at higher-than-normal presentation levels has been observed for listeners with and without hearing loss (Studebaker et al., 1999; Dubno et al., 2005; Molis and Summers, 2003; Shanks et al., 2002). Speech intelligibility predictors based on the acoustic signal properties, such as the articulation index and speech transmission index, cannot directly account for the effects of presentation level and hearing impairment. Recently, Elhilali et al. (2003) introduced the spectro-temporal modulation index (STMI), a speech intelligibility predictor based on a model of how the auditory cortex analyzes the joint spectro-temporal modulations present in speech. However, the auditory-periphery model used by Elhilali et al. is very simple and cannot describe many of the nonlinear, level-dependent properties of cochlear processing, nor the effect of hair cell impairment on this processing. In this study, we quantify the effects of speech presentation level and cochlear impairment on speech intelligibility using the STMI with a more physiologically-accurate model of the normal and impaired auditory periphery developed by Zilany and Bruce (2006). This model can accurately represent the auditory-nerve responses to a wide variety of stimuli across a range of characteristic frequencies and intensities spanning the dynamic range of hearing. In addition, outer and inner hair cell impairment can be incorporated. Compared to experimental word recognition scores, this model-based STMI can qualitatively predict the effect of presentation levels on speech intelligibility for both normal and impaired listeners in a wide variety of conditions