Short-term temporal integration: Evidence for the influence of peripheral compression

Abstract

Thresholds for a 6.5-kHz sinusoidal signal, temporally centered in a 400-ms broadband-noise masker, were measured as a function of signal duration for normally hearing listeners and listeners with cochlear hearing loss over a range of masker levels. For the normally hearing listeners, the slope of the function relating signal threshold to signal duration ͑integration function͒ was steeper at medium masker levels than at low or high levels by a factor of nearly 2, for signal durations between 2 and 10 ms, while no significant effect of level was found for signal durations of 20 ms and more. No effect of stimulus level was found for the hearing-impaired listeners at any signal duration. For signal durations greater than 10 ms, consistent with many previous studies, the slope of the integration function was shallower for the hearing-impaired listeners than for the normally hearing listeners. However, for shorter durations, there was no significant difference in slope between the results from the hearing-impaired listeners and those from the normally hearing listeners in the high-and low-level masker conditions. A model incorporating a compressive nonlinearity, representing the effect of basilar-membrane ͑BM͒ compression, and a short-term temporal integrator, postulated to be a more central process, can account well for changes in the short-term integration function with level, if it is assumed that the compression is greater at medium levels than at low or high levels by a factor of about 4. This is in reasonable agreement with physiological measurements of BM compression, and with previous psychophysical estimates.

Monaural temporal integration and temporally selective listening in children and adults

“…In other words, thresholds improved approximately linearly as the duration was incrementally lengthened by a factor of 4. Similar to the results reported by Oxenham et al ͑1997͒, the integration functions are steeper at shorter durations than at longer ones for all listener groups, a trend that is especially evident at 6500 Hz. Comparing results shown in Fig.…”

“…Symbols indicate data for three observer groups, including adults ͑open square symbols͒, children older than 7 years ͑open circles͒, and children younger than 7 years ͑filled circles͒. Despite differences in stimuli and experimental procedures, signal detection thresholds obtained from adult listeners at 6500 Hz are generally consistent with the data reported by Oxenham et al ͑1997͒. In all three groups of listeners signal detection thresholds decreased with increasing signal duration up to the maximum duration tested.…”

“…The particular approach used here is similar to that of Oxenham et al ͑1997͒, where a 6500-Hz signal that is brief with respect to the monaural temporal window can be used without resulting in substantial spectral splatter across frequency channels. In this paradigm, thresholds for a 6500-Hz signal are determined for a range of durations from 2 to 128 ms, measured from the half-rise point.…”

A computational model of human auditory signal processing and perception

“…The error bars indicate Ϯ one standard deviation across subjects, which is typically less than 1 dB but amounts to about 2 dB for the shortest signal duration of 5 ms. For signal durations in the range from 5 to 20 ms, the threshold decreases by about 4 -5 dB per doubling of signal duration, while the decrease is about 3 dB per doubling for durations above 20 ms. The data are consistent with results from earlier studies on signal integration in tone-in-noise masking ͑e.g., Dau et al, 1996b;Oxenham et al, 1997; The simulations ͑filled circles͒ show a constant decay of 3 dB per doubling of signal duration. This agrees nicely with the measured data for durations at and above 15 ms. At signal durations of 200 ms and above ͑not shown͒, the simulations are consistent with the prediction of a threshold of 48 dB obtained with the classical power spectrum model of masking ͑Patterson and Moore, 1986͒, assuming a threshold criterion of 1.5 dB increase in power ͑due to the addition of the signal to the noise͒ in the passband of the 2 kHz gammatone filter.…”

“…Each filter is followed by a nonlinear device. In recent models, the nonlinear device typically includes two processes, half-wave rectification and a compressive nonlinearity, resembling the compressive input-output function on the BM ͑e.g., Ruggero and Rich, 1991;Oxenham and Moore, 1994;Oxenham et al, 1997;Plack and Oxenham, 1998;Plack et al, 2002͒. The output is fed to a smoothing device implemented as a lowpass filter ͑Viemeister, 1979͒ or a sliding temporal integrator ͑e.g., Moore et al, 1988͒.…”