A procedure for testing speech intelligibility in a virtual listening environment

Objective: The development of a test of virtual speech intelligibility in noise that enables assessment in typical, everyday listening situations. To eliminate extraneous confounding factors, digital signal processing was incorporated to simulate listening environments and source locations and allow presentation of stimuli via earphones. Design: Source-to-eardrum transfer functions measured on KEMAR for various source locations in anechoic and reverberant environments were used to process monosyllabic words and speech-spectrum noise. Speech intelligibility was measured for three speech and noise configurations in two environments using an adaptive procedure to determine the signal-to-noise (S/N) ratio for 50% intelligibility. Results: Normal-hearing listeners achieved 50% intelligibility of monosyllabic words at significantly lower S/N ratios in a virtual anechoic environment than in a virtual reverberant environment. Speech intelligibility improved significantly in both environments when the speech and noise sources were separated, but the intelligibility gain in the anechoic environment was four times larger than in the reverberant environment. Conclusions: This test is easy to administer and score, and it provides a means for measuring: 1) the effects of separating speech and noise sources and 2) the effects of reverberation on speech intelligibility in noise while eliminating confounding factors such as calibration.