Contribution of Low-Level Acoustic and Higher-Level Lexical-Semantic Cues to Speech Recognition in Noise and Reverberation

Masking noise and reverberation strongly influence speech intelligibility decrease listening comfort. To optimize acoustics for ensuring a comfortable environment, it is crucial to understand the respective contribution of bottom-up signal-driven cues top-down linguistic-semantic recognition in reverberation. Since relevance these differs across test materials training status listeners, we investigate material type on noise, combinations We also examine performance subset measurement conditions. Speech measured with an open-set, everyday Plomp-type sentence compared scores closed-set Matrix-type consisting syntactically fixed semantically unpredictable sentences (c.f. data by Rennies et al., J. Acoust. Soc. America, 2014, 136, 2642–2653). While both tests yield approximately same threshold trained normal-hearing their may differ as result cognitive factors, i.e., more sensitive effects while open-set affected language familiarity. All experimental were obtained at signal-to-noise ratio (SNR) and/or time set obtain desired transmission index (STI) values 0.17, 0.30, 0.43. respectively, thus linking STI predictions measure pure low-level acoustic effects. The results confirm consistent difference between robustness observed literature matrix sentences, especially poor medium intelligibility. against disappeared when listeners had no priori knowledge about (sentence structure words used), demonstrating higher-level lexical-semantic recognition. In addition, reverberation- noise-induced high conditions differences clearly demonstrate limited utility predicting