Auditory Model-Based Sound Direction Estimation With Bilateral Cochlear Implants.

Area of Science:

• Auditory Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Bilateral cochlear implants (CIs) improve sound localization compared to single CIs, but performance remains suboptimal.
• Previous research focused on interaural level and time differences in the temporal envelope.

Purpose of the Study:

• To develop and validate a binaural model predicting sound source localization in the azimuthal plane for bilateral CI users.
• To investigate the relationship between model predictions and experimental data.

Main Methods:

• A two-channel binaural model was created, incorporating a clinical speech-coding strategy and neural interface models.
• The model included prediction stages trained to map neural response rates to azimuthal angles.
• Virtual acoustics were used to generate diverse noise and speech stimuli for training and testing.

Main Results:

• The model successfully predicted azimuthal sound direction based on bilateral CI input.
• Localization error patterns generated by the model closely matched experimental data from bilateral CI users.
• The model's performance was largely explained by the nonmonotonic relationship between interaural level difference and azimuthal angle.

Conclusions:

• The developed binaural model provides a robust framework for understanding sound localization in bilateral CI users.
• The findings highlight the importance of interaural level differences and their complex relationship with azimuthal angle in CI sound perception.