短期統計學習緩解聲音定位的不適定問題

// Short-Term Statistical Learning Mitigates the Ill-Posed Problem of Sound Localization

本研究探討聽者能否在短時間內藉由學習聲源穩定的頻譜特性來改善聲音定位能力｜實驗發現，預測性頻譜可在數分鐘內將重大定位錯誤降低最多5%｜結果顯示聽者會根據近期聲音經驗調整空間聽覺期望，展現聽覺空間推理的高度彈性與適應性。

// This study investigated whether listeners can improve sound localization by learning a sound source's consistent spectral pattern over a short time. Participants localized sounds with either fixed or randomized spectral patterns, and predictable spectra reduced major localization errors by up to 5% within minutes. The results show that listeners adapt their spatial hearing by updating spectral expectations based on recent acoustic experience. This highlights the flexibility and adaptive nature of auditory spatial inference.

Published 2026年7月9日