Fig 1.
Experimental design and analysis framework.
(A) Location of loudspeakers (for this particular example, S1: Target speech, S2: Masker speech, B1-B4: Babble noise) relative to the participant. (B) Timeline of each experimental trial. Start of the noise babble at t0 = 0, followed by target and masker at t1 = 5s. (C) Neural signals were recorded with EEG as participants listened to natural speech monologues. A lagged linear regression analysis was carried out to estimate the Temporal Response Function (TRF) describing the relationship between speech phonetic features (S—Spectrogram, F -Phonetic Features, D—Half-wave rectified Spectrogram Derivative) and low-frequency EEG (1–8 Hz).
Fig 2.
Stronger cortical encoding of target acoustic-phonetic information compared to masker information in listeners with hearing impairment.
(A) EEG prediction correlations (Pearson’s r) for phonetic features models F and shuffled control Fsh of the target and masker speech. Boxplots show the median and inter-quartile range (IQR) of the distributions. Scalp topographies represent the distribution of prediction correlations across all channels (average across participants) for the F model. (B) TRF weights at channel FCz for the eighteen phonetic features for the F model. (C) Phoneme distance maps (PDMs) for target (red) and masker (green) speech. (D) EEG sensitivity to groups of phonetic features, i.e., quality of clustering of the EEG responses around relevant phonetic contrasts, for target and masker speech.
Fig 3.
Representation of phonemic onsets of the ignored speech sounds in the low-frequency EEG of participants with HI.
(A) EEG prediction gains obtained from the PhCat (FS-S) and PhOnset (FshS-S) metric, for the target and masker speech. Bars represent the increase in prediction correlations (r) averaged across all participants and electrodes. Error bars represent the SEM across participants. (B) Topographical distribution of the average EEG prediction correlation increases from the baseline model S, across all electrode locations.