Z. Peng, L. M. Wang, and S. K. Lau. (2012) in Proceedings of InterNoise12, New York City, NY, pp.1865-1875.
This study investigates the effects of acoustic conditions on speech comprehension, rather than speech intelligibility as often reported in existing literature. Sets of 15-minute-long listening comprehension tests were developed based on the format of the Test of English for International Communication (TOEIC). Each test set includes four types of tasks: matching aural phrases to photographs, selecting appropriate responses to aural questions, and answering questions after listening to conversations (between two talkers) and talks (single talker). Within the Nebraska acoustics test chamber, native-English-speaking participants are asked to perform these tests under 15 acoustic conditions, from combinations of three background noise levels (RC-30, 40 and 50) and five mid-frequency reverberation times (0.4 to 1.2 seconds). The background noise levels are varied via an Armstrong i-Ceiling system, while the reverberation times are simulated from convolving the anechoic test signals with binaural room impulse responses (BRIR), simulated in ODEON for a typical classroom. A two-channel playback system is used to present the convolved audio signals, with loudspeaker-listener configuration embedded in the BRIR auralization output. Pilot testing of three subjects showed no variation of performance scores on overall tasks among all acoustical conditions. However, participants generally scored lowest in tasks to comprehend conversations in the longest RT scenarios.
Available at: http://works.bepress.com/lilymwang/20/