Brian K. Man1, Elaine H. N. Ng1,2, Emina Alickovic3,4
1Oticon A/S, Smørum, Denmark
2Department of Behavioral Sciences and Learning, Linkoping University, Linkoping, Sweden
3Eriksholm Research Centre, Oticon A/S, Snekkersten, Denmark
4Department of Electrical Engineering, Linkoping University, Linkoping, Sweden

Everyday listening situations that require listeners to selectively attend to a talker of interest in noisy environments with multiple competing talkers are among the most challenging situations encountered by hearing-impaired listeners. Such challenges become even more pronounced with increasing background noise level and may partially be overcome by adequate hearing aid (HA) amplification and noise reduction (NR) support. Using Electroencephalography (EEG), it has been demonstrated that the auditory cortex selectively represents the target talker with significantly higher fidelity than other competing talkers in normal-hearing and hearing-impaired listeners. An NR scheme in commercial HAs was also found to enhance the neural representation of the foreground and entire acoustic scene in early (<85 ms) EEG responses and to enhance neural representations of the target and masker speech and suppress the neural representation of the background noise in late (> 85ms) EEG responses. Motivated by these findings, in this review we investigate whether neural representation of speech in distinct hierarchical stages is affected by NR scheme and configuration in commercial HAs. We show that both the choice of NR scheme and configuration can significantly affect the neural speech processing. By using auditory-attention decoding methods, we show that selecting the adequate NR configuration leads to significantly better representation of the entire acoustic scene and foreground in early EEG responses and to significantly better representation of target and masker speech in late EEG responses in noisy environments. These results suggest that EEG-based auditory attention decoding methods may be sensitive to the choice of HA signal processing configurations in HA users.