Dynamic Reweighting of Auditory Modulation Filters

Abstract

Sound waveforms convey information largely via amplitude modulations (AM). A large body of experimental evidence has provided support for a modulation (bandpass) filterbank. Details of this model have varied over time partly reflecting different experimental conditions and diverse datasets from distinct task strategies, contributing uncertainty to the bandwidth measurements and leaving important issues unresolved. We adopt here a solely data-driven measurement approach in which we first demonstrate how different models can be subsumed within a common ‘cascade’ framework, and then proceed to characterize the cascade via system identification analysis using a single stimulus/task specification and hence stable task rules largely unconstrained by any model or parameters. Observers were required to detect a brief change in level superimposed onto random level changes that served as AM noise; the relationship between trial-by-trial noisy fluctuations and corresponding human responses enables targeted identification of distinct cascade elements. The resulting measurements exhibit a dynamic complex picture in which human perception of auditory modulations appears adaptive in nature, evolving from an initial lowpass to bandpass modes (with broad tuning, Q∼1) following repeated stimulus exposure.

Mechanisms of spectrotemporal modulation detection for normal- and hearing-impaired listeners

“…This could be a first reason explaining why we did not observe any learning between the beginning and the end of the experiment. Note that previous studies using masking modulation paradigms found plasticity for AM-processing (Joosten et al, 2016) but not with STM-processing (Oetjen & Verhey, 2015). Second, it is possible that the learning occurred extremely rapidly during the initial training phase or pre-test (see Grant et al 2013) and could therefore not be exposed here, since our pre- and post-tests were scheduled at the end of the training session and the final session, respectively.…”

“…These filters show different patterns for target-absent vs. target-present stimuli in both groups, exposing the presence of non-linear processes unaccounted for by a template-matching strategy (Neri, 2004; Tjan & Nandy, 2006; Neri, 2010b). These departures from template-matching are accommodated by a small cascade model consisting of a front-end STM weighting function followed by a MAX operation (see also Joosten et al, 2016). Because this cascade structure is applicable to both NH and HI groups, this result indicates that hearing-impaired listeners rely on intact circuitry for monitoring the output of their modulation channels.…”

“…We used reverse-correlation to derive “;perceptual filters” engaged by listeners in our task (Murray, 2011). Filters were derived from each individual separately for “target-absent” and “target-present” stimuli to assess the presence of potential nonlinear processes: if listeners behave like template matchers, target-absent and target-present perceptual filters must match; if they differ, this result indicates the presence of nonlinear strategies that cannot be subsumed under the template-matching operation (Joosten et al 2016). Using the notation introduced earlier and in keeping with current literature, the target-absent perceptual filter was computed as p [0] =avg( n i [0,0] )-avg( n i [0,0] ) while the target-present filter was computed as p [1] =avg( n i [1,1] )-avg( n i [1,0] ), where avg() indicates averaging across all trials of the indexed type.…”

“…Because the above experiments involved NH listeners, we do not know how cochlear hearing loss impacts the characteristics of this process (tuning, directionality and possibly other characteristics). Most importantly, it is unknown how the auditory system makes use of different STM filters; the computational mechanisms that underpin the processing of spectral-only and temporal-only modulations have been explored in various studies (Dau et al 1997; Ewert et al, 2002; Joosten et al, 2016), but the psychophysical mechanisms for joint spectral and temporal modulations have not been examined directly.…”

“…Previous studies relying on this technique have clarified various aspects of suprathreshold auditory tasks, e.g. nonlinear mechanisms for tone-in-noise detection (Joosten et al, 2012), retuning of AM-processing (Joosten et al, 2016), modulation primitives supporting speech-in-noise intelligibility (Venezia et al, 2016, 2019). Successful application of reverse-correlation relies, among others, on two inter-related factors: (i) the nature and structure of the perturbing noise source must efficiently interfere with the mechanisms engaged by listeners for detecting the target, and (ii) a large data mass (several thousand trials) is necessary to derive a stable, accurate image of those mechanisms.…”