Spatial and Spectral Factors in Release from Informational Masking in Speech Recognition

When the speech of one talker is presented within a background of one or two other talkers, under some conditions there appears to be more interference than one might expect based on theories of traditional "energetic" masking. The current studies investigated the manner in which this type of additional "informational" masking can be overcome and how release from informational and energetic masking may differ. The target sentences were spoken by a female talker, and the masker was either the speech of two additional female talkers or continuous speech-spectrum noise. In the first of two experiments, the target was presented from a front loudspeaker and the masker was presented either from the front or from directly behind the listener. Release from masking in the front-back condition was greater for the two-talker masker than for the noise masker. In the second experiment, both target and maskers were high-pass filtered at 715 Hz, and energy below 215 Hz was added back to the target. When spatial differences were eliminated, the low-frequency energy available in the target was considerably more beneficial when the target was presented against the two-talker masker than against the noise background. The overall results suggest that spatial and spectral differences lead to perceptual distinctions that are helpful to the listener in tracking the target message in the presence of other voices.