Objective: A generalised concept for hearing aid fitting and fine-tuning based on estimated individual traits is presented along first implementations in this report.
Design: To estimate the individual traits, a set of auditory model-based performance measures is used to generate promising candidates within the algorithm's parameter space for a subsequent subjective rating. For the subjective assessment, a fast and intuitive multi-stimulus test denoted as combined discrimination and classification (CoDiCl) is presented to capture user preferences for an optimised setting.
Study sample: The estimation of individual traits is shown in an exemplary manner for a multidimensional coherence-based noise reduction algorithm. The dimensionality reduction was performed using differently weighted combinations of speech intelligibility index (SII) and perceived similarity measure (PSM).
Results: Nine reasonable alternative algorithm setting candidates were extracted from a model-optimised exploration path (MOEP) for a subsequent subjective rating to potentially differentiate between listeners with different attitudes towards noise suppression and introduced distortions (i.e. "noise haters" and "distortion haters").
Conclusions: By iteratively improving the agreement between subjective and objective assessment, an objective estimation of subjective traits using appropriate weightings of objective measures may become possible. This will potentially help to efficiently fit modern multidimensional hearing aid algorithms to the individual user.
Keywords: Hearing aid fitting; candidature criteria; dimensionality reduction; fine-tuning; individualisation; meta-controller; subjective evaluation.