Speech intelligibility and spatial release from masking in a vehicle cabin

Spatial separation of the target and masker usually improves intelligibility (i.e., spatial unmasking), but this is not well understood in vehicle cabins. This study measured how target-masker separation in a stationary vehicle cabin affects intelligibility and compared spatial release from masking (SRM) across three conditions: a vehicle with the left front window open or closed and a weakly reflective listening room. The listener was always in the driver’s seat; target and masker came from different passenger seats in co-located and separated configurations, yielding 12 target-masker configurations (3 target positions × 4 masker positions) for each acoustic condition. The masker was speech-shaped noise to isolate energetic masking and assess spatial unmasking from acoustic cues alone. Stimuli were created by convolving signals with binaural impulse responses measured at the driver’s position using a dummy head. Mandarin speech reception thresholds (SRTs) were measured over headphones (N=10), and SRM was derived from these SRTs. A binaural speech intelligibility model was used for further SRM analysis. Results show that for a driver-seat listener, speech from the front-passenger seat is most intelligible and least affected by other-seat maskers, but it is also the most disruptive when it serves as the masker. The back-right seat yields the poorest intelligibility yet causes the least interference with targets at other seats. In the listening room, larger target-masker separation generally increases SRM. In the vehicle cabin, SRM depends on both spatial separation and the non-uniform effects of seat occlusion and boundary reflections. Because the vehicle’s sound field is non-uniform, interaural correlation is reduced and binaural advantage is weakened. SRM in the vehicle is therefore dominated by the better-ear effect, with binaural unmasking contributing less than about 1.5 dB. Opening the left front window has only a small impact on overall intelligibility and SRM but can slightly enhance binaural unmasking by partially restoring binaural advantage.