Abstract:
A multi-source listening environment was simulated under headphones by recording signals through KEMAR in a sound-deadened room. Seven loudspeakers were each positioned in the frontal plane 5 ft from KEMAR at 30-deg increments, ranging from +90 (right) to -90 (left). The recorded signals (IEEE Harvard sentences) were digitally mixed and played back under headphones in a sound-proof booth using a Digital Audio Tape (DAT) player. Sound locazation accuracy was measured for ``target'' sentence in the presence of various ``jammers.'' Three variables were manipulated. First, the number of ``jammers'' was either none, one, two or three. Second, the relative location of the targets and jammers was varied, such that they were either in close proximity, widely separated, or some of each. Third, the jammer content comprised of either other sentences, speech-shaped noise, babble or time-reversed sentences. Normal-hearing listeners were tested under binaural and monaural conditions. The content of the target sentence was known. Localization accuracy decreased as the number of jammers increased, and as the proximity of the target and jammers was increased. In addition, performance was worse under monaural than under binaural conditions. [Work supported by NIH DC02696, DC00100-21.]