AbstractRegions of the human posterior superior temporal gyrus and sulcus (pSTG/S) respond to the visual mouth movements that constitute visual speech and the auditory vocalizations that constitute auditory speech. We hypothesized that these multisensory responses in pSTG/S underlie the observation that comprehension of noisy auditory speech is improved when it is accompanied by visual speech. To test this idea, we presented audiovisual sentences that contained either a clear auditory component or a noisy auditory component while measuring brain activity using BOLD fMRI. Participants reported the intelligibility of the speech on each trial with a button press. Perceptually, adding visual speech to noisy auditory sentences rendered them much more intelligible. Post-hoc trial sorting was used to examine brain activations during noisy sentences that were more or less intelligible, focusing on multisensory speech regions in the pSTG/S identified with an independent visual speech localizer. Univariate analysis showed that less intelligible noisy audiovisual sentences evoked a weaker BOLD response, while more intelligible sentences evoked a stronger BOLD response that was indistinguishable from clear sentences. To better understand these differences, we conducted a multivariate representational similarity analysis. The pattern of response for intelligible noisy audiovisual sentences was more similar to the pattern for clear sentences, while the response pattern for unintelligible noisy sentences was less similar. These results show that for both univariate and multivariate analyses, successful integration of visual and noisy auditory speech normalizes responses in pSTG/S, providing evidence that multisensory subregions of pSTG/S are responsible for the perceptual benefit of visual speech.Significance StatementEnabling social interactions, including the production and perception of speech, is a key function of the human brain. Speech perception is a complex computational problem that the brain solves using both visual information from the talker’s facial movements and auditory information from the talker’s voice. Visual speech information is particularly important under noisy listening conditions when auditory speech is difficult or impossible to understand alone Regions of the human cortex in posterior superior temporal lobe respond to the visual mouth movements that constitute visual speech and the auditory vocalizations that constitute auditory speech. We show that the pattern of activity in cortex reflects the successful multisensory integration of auditory and visual speech information in the service of perception.
Using Visual Speech Information in Masking Methods for Audio Speaker Separation
