Effects of non-individual head-related transfer functions and visual mismatch on speech recognition in virtual acoustic environments

Abstract

There is widespread research and clinical interest in quantifying how the acoustics of real-world environments, such as background noise and reverberation, impede a listener’s ability to recognize speech. Conventional methods used to quantify these effects include dichotic listening via headphones in sound-attenuated booths or loudspeakers in anechoic or low-reverberant environments, which lack the capability of manipulating room acoustics. Using a state-of-the-art Variable Room Acoustics System housed in a virtual sound room (ViSoR), this study aims to systematically assess the effects of non-individual head-related transfer functions (HRTFs) and mismatched visual perception on speech recognition in virtual acoustic environments. Young adults listened to and repeated sentences presented amidst a co-located two-talker speech competitor with reverberation times ranging from 0.4 to 1.25 s. Sentences were presented in three listening conditions: through a loudspeaker array in ViSoR with the participants’ own HRTFs (Condition 1); via headphones in a sound-attenuated booth with non-individual HRTFs(Condition 2); and using the same binaural reproduction as Condition 2 in ViSoR (Condition 3). Condition 3 serves as a control condition, allowing us to quantify the separate effects of non-individual HRTFs and visual mismatch on speech recognition. Discussion will address the validity and use of virtual acoustics in research and clinical settings.