利用流形學與內在測度進行聽覺特徵之客觀比較

// Objective Comparison of Auditory Profiles Using Manifold Learning and Intrinsic Measures

本研究探討分群方法及群組數量對聽力受損者聽覺輪廓生成的影響｜比較8種既有輪廓框架，並運用統計指標與流形學習分析大型公開資料集｜結果顯示分群策略及群組數顯著影響輪廓品質，Hearing4All框架在區分度指標表現最佳，具備未來應用潛力。

// This study examined how clustering methods and the number of profiles affect auditory profile generation for individuals with hearing impairment. Eight established profiling frameworks were compared using statistical measures and manifold learning on a large open-access dataset. Results showed that both clustering approach and profile count significantly impact profile quality, with the Hearing4All framework performing best on separability measures. The study highlights the importance of clear group differentiation and identifies Hearing4All as a promising method for future auditory profiling.

Published 2026年7月1日