Authors: Jack Webb¹, Christophe Lesimple², Volker Kuehnel², Lorenzo Picinali¹
¹Imperial College London; ²Sonova AG
Background: Music quality in hearing aids has traditionally been investigated as a monaural percept, often overlooking the influence of distortions to interaural cues induced by bilateral processing. While disruptions to interaural time, level, and coherence cues are known to impair auditory scene analysis, their specific impact on the perceived quality of music remains underexplored.
Method: Twenty normal-hearing participants assessed 268 music signals processed with a hearing aid and hearing loss simulator using an adaptive pairwise comparison paradigm. Perceived quality scores were modeled using features capturing both monaural magnitude spectral distortions and interaural cue distortions. A neural model was additionally trained to predict quality from these combined feature sets.
Results: Regression analysis revealed that binaural cue distortions were associated with a statistically significant decrease in perceived quality, independent of monaural spectral distortions. While the neural model outperformed several existing metrics, it did not exceed the performance of the best monaural model when averaged across channels, suggesting a relatively small effect size for interaural distortions. Results further suggest that the weighting of these distortions may depend on the signal’s acoustic characteristics.
Conclusion: These findings indicate that binaural cues contribute to music quality in simulated hearing aid contexts. Future hearing aid signal processing strategies may benefit from preserving interaural cues and accounting for the complex interactions between processing algorithms and the specific acoustic nature of the musical input.
