FIG. 3.

(Color online) Our rigorous two-step approach to test the hypothesis that the fidelity of neural envelope coding of target speech relative to background noise predicts speech intelligibility (in a quantitative, statistical sense). The first step is a calibration step, where a logistic/sigmoid function was used to map an EEG-based target envelope-coding metric, ENVneural, to perceptual intelligibility for speech in stationary noise. In the second step, we used this mapping to blindly predict speech intelligibility in various completely novel realistic background noises and distortions only from EEG-based ENVneural measurements.