FIGURE 7.

(H2) The NR algorithm enhances the neural representation of the ignored speech to a higher degree at a lower signal-to-noise ratio (SNR) level than at a higher SNR. Boxplot showing the reconstruction accuracy value of the ignored speech envelope reconstructed from the ignored talker decoder for two SNR levels tested with noise reduction (NR) scheme inactive (OFF) and active (ON; left column) for global averages across NR schemes (middle column) and for global averages across SNR levels (right column). A significant main effect of NR, indicating the enhanced neural representation of the ignored speech when NR scheme was turned on, and a significant main effect of SNR were observed. Significant differences in reconstruction performances were seen between inactive and active NR in the pairwise comparison at lower SNR (+3 dB). The horizontal lines in yellow denote single participants and the horizontal lines in black depict predictions of linear mixed models. The black horizontal line in the boxplot denotes the mean reconstruction accuracy of the ignored speech envelope and the box indicates the upper and the lower quartiles, with the vertical lines representing the minimum and the maximum reconstruction accuracy values. The asterisk indicates significant differences (**p < 0.01, ****p < 0.0005); ns, not significant.