Figure 9: eSTRF changes overlap with speech features.
From: Rapid tuning shifts in human auditory cortex enhance speech intelligibility
Top: spectrograms of unfiltered speech (a) were convolved with the eSTRF fit on each filtered speech condition (b). The size of the predicted response (c, left) depends on the overlap between the eSTRF and the unfiltered speech features. Scatterplot (c, right) shows the predicted response power between BEFORE and AFTER conditions. Size and colour represent the difference (AFTER–BEFORE), and inset histogram shows the distribution of differences. Bottom: spectrograms of filtered speech (d) were convolved with the eSTRF fit on unfiltered speech (e), resulting in a predicted HFB amplitude for that spectrogram (f, left side, grey trace). This was compared with the true HFB activity in each condition (green and orange traces). Correlation between the predicted and actual trace reflects the extent to which an unfiltered eSTRF is predictive of the neural response. The scatterplot shows the comparison between Predicted and True AFTER versus Predicted and True BEFORE (f, right). Size and colour of points represent the difference (AFTER–BEFORE), and inset histogram shows the distribution of this difference.
