Modeling binaural speech intelligibility in symmetrical listening conditions

Modeling speech intelligibility in spatially separated versus co-located masker conditions reveals how listeners use spatial cues, such as interaural time and level differences, to separate speech from noise. Symmetric listening conditions, with maskers symmetrically positioned around a frontal target,are particularly interesting because they do not offer a long-term signal-to-noise advantage at either ear. Binaural advantages arise solely from binaural unmasking (BU) and short-term better-ear listening (BE). We used the binaural speech-based envelope power spectrum model to evaluate spatial release from masking (SRM) in symmetric two-masker conditions. The model analyzes amplitude modulations of noise and noisy speech using short-time windows. It includes two monaural paths—one per ear—to simulate BE, and a binaural path that integrates information across ears with an equalization-cancellation process, simulating BU. SRM predictions were compared with existing behavioral data across different spectro-temporal maskers in anechoic and reverberant symmetrical conditions. While the model captures the masker-dependent trends, it overestimates SRM in modulated maskers. The BE path alone accurately reproduces SRM across masker types, but the model's binaural selector favors the BU path, leading to an SRM overestimation. We analyze independent BE and BU simulations, evaluate their contributions, and compare the model to other binaural models.