Effects of spectral smearing on performance of the spectral ripple and spectro-temporal ripple tests

Abstract

The main aim of this study was to use spectral smearing to evaluate the efficacy of a spectral ripple test (SRt) using stationary sounds and a recent variant with gliding ripples called the spectro-temporal ripple test (STRt) in measuring reduced spectral resolution. In experiment 1 the highest detectable ripple density was measured using four amounts of spectral smearing (unsmeared, mild, moderate, and severe). The thresholds worsened with increasing smearing and were similar for the SRt and the STRt across the three conditions with smearing. For unsmeared stimuli, thresholds were significantly higher (better) for the STRt than for the SRt. An amplitude fluctuation at the outputs of simulated (gammatone) auditory filters centered above 6400 Hz was identified as providing a potential detection cue for the STRt stimuli. Experiment 2 used notched noise with energy below and above the passband of the SRt and STRt stimuli to reduce confounding cues in the STRt. Thresholds were almost identical for the STRt and SRt for both unsmeared and smeared stimuli, indicating that the confounding cue for the STRt was eliminated by the notched noise. Thresholds obtained with notched noise present could be predicted reasonably accurately using an excitation-pattern model.

Effect of ripple repetition rate on discrimination of ripple glide direction and the detection of brief tones in spectro-temporal ripple noise

“…Bonferroni-corrected pairwise comparisons showed that thresholds were significantly lower (worse) for R ¼ 16 and 32 Hz than for all other values of R. The value of D at threshold was just above 5 RPO for R 8 Hz, consistent with the results of Narne et al (2018) for R ¼ 5 Hz. The median threshold across participants for the condition with static ripples was similar to that reported by Narne et al (2016) and was also similar to the median threshold across participants for the conditions with R ¼ 2-8 Hz.…”

“…The STRn was generated exactly as described by Narne et al (2016). The carrier was composed of 201 equal-amplitude sinusoidal frequency components, spaced every 0.03 octave from 100 to 6400 Hz.…”

“…As a result of the temporal changes, the overall loudness evoked in each local frequency region, the energy at the spectral edges, and the spectral centroid became unreliable cues for discriminating the target and reference stimuli. However, Narne et al (2016) showed that the results for the STRt could be affected by a confounding cue, namely, amplitude fluctuations with rate R at the outputs of auditory filters tuned above and below the passband of the stimuli, which occur even when R is sufficiently high that such fluctuations are negligible at the outputs of auditory filters tuned within the passband of the STRt.…”

“…A variant of the STRt is the STRt dir described by Narne et al (2018), which estimates the highest value of D at which an upward-gliding ripple can be discriminated from a downward-gliding ripple. This test was intended to avoid both the confounding cues described by Azadpour and McKay (2012) for stimuli with static spectral ripples and the confounding cues described by Narne et al (2016) for stimuli with gliding spectral ripples. The cues become less usable in the latter case because the upward-and downward-gliding ripples exhibit identical long-term modulation spectra in frequency regions near the lower and upper edges of the passband of the stimuli.…”

“…For very high rates (>1000 Hz), the AM is essentially undetectable when the carrier frequency is high and the spectral sidebands are not resolved, although performance can improve with increasing AM rate when the sidebands are resolved (Kohlrausch et al, 2000;Moore and Glasberg, 2001). In the case of the STRt, we are not aware of any studies assessing the effect of R. Following Aronoff and Landsberger (2013), several studies have used the STRt with R ¼ 5 (Kirby et al, 2015;Narne et al, 2016;Vickers et al, 2016;Zhou, 2017). However, no clear rationale was given for the choice of R, and it is not clear whether temporal resolution influences the detectability of spectral ripples for this value of R.…”