Mean standard deviation for the NH and HI listener groups.
Long-term rms power spectra (solid curves) for the spectrally flattened target speech and the associated dynamic range (shaded areas) presented to NH at 57 SPL (left panel) and to HI listeners at 87 SPL (right panel). The original speech power spectra before spectral shaping (dotted curves) are shown for comparison. Target signal audibility in quiet is represented by the proportion of the shaded area falling above the absolute hearing threshold (dashed curves).
Mean performance in identifying sentence keywords for target speech presented in three types of masker as a function of SNR. Symbols indicate the experimental data, with error bars standard error across listeners. Solid curves indicate sigmoid fits to the stationary-noise data. Dashed and dotted curves indicate ESII predictions for the modulated-noise and interfering-talker conditions, respectively, calculated using the SNR-dependent fluctuating-masker IIFs depicted in Fig. 8.
(Upper panel) Mean SRTs are plotted for each listener group, modality, and masker. (Lower panel) The mean FMB, defined as the difference between the stationary and fluctuating-masker SRTs, is plotted for each listener group, modality, and fluctuating masker. Error bars standard error across listeners.
Experimental results for individual listeners (symbols), showing an inverse relationship between the magnitude of the FMB and the stationary noise SRT across listener groups and audio-alone and AV presentation modalities. Individual curves indicate a similar inverse relationship across a range of stationary-noise SNRs within each listener group and modality. The vertical separation between pairs of curves indicates FMB differences that remain once baseline stationary-noise SNR differences were controlled.
A schematic drawing describing how a SNR-dependent FMB might arise. The functions in each panel represent an interpolated version of the IIF derived by Studebaker and Sherbecoe (2002), indicating the distribution of speech information across the dynamic range. Fluctuating maskers contain both peaks and valleys, which mask (light shaded area) and unmask (dark shaded area) speech information relative to the stationary-noise condition. For an average SNR of (left panel), the masking and unmasking of speech information are approximately equal, yielding no net FMB. For an average SNR of (right panel) the amount of speech information unmasked in the masker valleys is greater than the information masked by the masker peaks, yielding a net FMB benefit.
ESII model FMB predictions as a function of the stationary-noise SRT, with audiometric thresholds and signal levels set to represent the average NH (dashed curves) and HI (solid curves) listener, with three assumed IIFs: (thin curves) a uniform distribution (ANSI, 1997; Rhebergen et al., 2006); (medium curves) the Studebaker and Sherbecoe (2002) IIF estimate; and (thick curves) a SNR-dependent fluctuating-masker IIF fit to the experimental data. Data for individual listeners (symbols) are replotted from Fig. 5.
The Studebaker and Sherbecoe (2002) stationary-noise IIF (dashed curve) and a series of SNR-dependent fluctuating-masker IIFs (solid curves) used in the ESII simulations to produce the FMB predictions of Fig. 7.
Article metrics loading...
Full text loading...