Estimating cochlear-filter shapes, temporal-window width and compression from tone-sweep detection in spectral and temporal noise gaps

Abstract

A test, designed for naive listeners, measured tone-sweep detection in noise with either spectral or temporal gaps. For normal-hearing (NH) listeners, detections in spectral gaps depended on level, which can be explained from Outer-Hair-Cell (OHC) activity. At low levels, OHC activity increased frequency-selectivity by amplifying the signal in the spectral gap, improving the signal-to-noise ratio (SNR). Relative to the broad passive cochlear filter, OHC activity decreased with rising signal levels. In consequence, SNRs decreased and detection deteriorated. Similarly, decreasing OHC activity may explain the observed level dependence of detection thresholds in temporal gaps. At low and high intensities, signal and noise were equally amplified. Detection was best at intermediate levels when the low-level signal in the temporal gap was amplified more than the high-level noise. All effects are modeled using a one-parameter time window with decaying-exponential shape preceded by a simplified dual-resonance non-linear (DRNL) filter. The filter contains two parallel, one-parameter, Rounded Exponential filters: a broad filter representing passive cochlear filtering and a narrow one, combined with a level-dependent amplifier, representing OHC activity. From estimated filter and time-window widths and OHC amplification, compression curves are derived. Additionally, results for hearing-impaired listeners will be presented.