The Role of Reverberation in Release from Masking Due to Spatial Separation of Sources for Speech Identification

Arbogast et al. [1] found a large release from masking obtained by the spatial separation of a target talker and a competing speech masker. Both signal and masker were sentences from the Coordinate Response Measure corpus processed by extracting the envelopes of 15 narrow frequency bands and using the envelopes to modulate carrier tones at the center of each band. By playing nonoverlapping subsets (6–8) of bands from the signal and masker, the energetic component of masking was minimized while the informational component was maximized. The current study extended that work to determine the interaction between reverberation, masker type and spatial release from masking. The stimuli were processed in the same way and were presented in the same spatial configuration as the earlier study. The target sentence was presented at 0-deg azimuth while the masker sentence was played at either 0- or 90-deg azimuth. Noise-masker controls, comprised of overlapping or nonoverlapping frequency bands, were also tested. The listening environment was an IAC booth having interior dimensions of 12'4" × 13' × 7'6". Acoustic extremes were achieved by covering all surfaces with materials that were either highly reflective (Plexiglas^® panels) or highly absorbent of sound (Silent Source^® foam wedges). The results indicated that the amount of masking and the spatial release from masking depended both on the characteristics of the room and the masker type. When the masker was primarily energetic, spatial release from masking decreased from a maximum of about 8 dB in the least reverberant room to about 2 dB in the most reverberant room. For the informational masker, a larger advantage of 15–17 dB was found that was not affected by reverberation. Our interpretation of these findings was that spatial separation of sources could improve speech identification through acoustic filtering by the head, binaural interaction, and the strengthening of perceptual segregation of sound images. However, only the latter effect appears to be relatively insensitive to reverberation.