The curve bisection procedure used to derive the RTF. Panel A denotes the value for 0.50 SII, panel B denotes the value for 0.25 SII, and panel C denotes the values used to derive 0.75 SII. Results from high-pass filtering conditions are represented by lines with upper ends that start from the left side of the graph; those from low-pass filtering conditions start from the right side.
Mean percent speech intelligibility, plotted as a function of cutoff frequency at various SNRs. Results from high-pass filtering conditions are represented by lines with upper ends that start from the left side of the graph; those from low-pass filtering conditions start from the right side.
PI function plotted as mean percent intelligibility at various SNRs (refer to individual RTSs). The bars represent standard error from the mean.
Crossover frequency as the intersection between regression lines as a function of mean percent intelligibility at cutoff frequencies from 500 to . Results from SNR (refer to individual RTSs) conditions are used.
Best-fit relative transfer function (RTF) and the 13 intelligibility scores (%) plotted as a function of SII values.
Frequency-importance function (FIF) of the CHINT.
Comparison of cumulative FIFs derived from the CHINT and other similar materials.
Best-fit absolute transfer function (ATF) and actual mean scores (%) plotted as a function of SII.
Crossover frequencies of various speech materials.
Description and examples of each Cantonese tone.
Mean percent speech recognition scores in various filtering/SNR conditions used in the actual experiment. The blank cells represent conditions that were not evaluated in the study.
Frequency-importance function in one-third octave bands. The weights are expressed as percentages (%).
Comparison of transfer functions (TFs) and frequency-importance functions (FIFs) for various speech materials.
Article metrics loading...
Full text loading...