Modeling binaural speech intelligibility in spatial listening conditions Free

Speech intelligibility is substantially improved when speech and interfering noise are spatially separated. This spatial unmasking is mostly caused by a combination of head shadow and binaural auditory processing. Binaural speech reception thresholds (SRTs) in such spatial conditions can be predicted very accurately using a combination of an Equalization-and-Cancellation (EC) model and the Speech-Intelligibility-Index (SII). This binaural speech intelligibility model predicts effects including levels, frequency spectra, and directions of the speech and noise signals as well as listeners' hearing loss, early reflections and reverberant parts of the noise signals. Earlier versions of the model were only able to predict the intelligibility of near-field speech. Recent extensions can also predict the intelligibility of far-field speech by taking early reflections and reverberant parts of the speech signal into account. However, some interactions between the direction of the noise source and early speech reflections cannot be predicted yet. The overall high prediction accuracy of the model (more than 90% of the data's variance can be explained) indicates that the model is applicable in real rooms and may serve as a tool in room acoustical design. This work was supported by the Deutsche Forschungsgemeinschaft (SFB TRR 31).