對噪聲聲碼聲音克隆的感知適應

// Perceptual Adaptation to Noise-Vocoded Voice Clones

本研究比較了噪聲聲碼器（重度語音扭曲技術）下，語音仿製聲與真人聲的可懂度｜結果顯示聽眾對仿製聲的理解度高出13.2%，且對兩種聲音的適應能力相當｜此發現指出仿製聲在嚴重訊號退化情境下仍具優勢，對輔助聽力科技有重要應用價值

// This study tested whether voice clones remain more intelligible than human voices when degraded by noise-vocoding, a method that heavily distorts speech. Ten noise-vocoded cloned voices and their human originals were compared in listener tests involving 80 sentences. Results showed that listeners understood cloned voices 13.2% better than human voices and adapted similarly to both types over time. These findings suggest cloned voices retain an intelligibility advantage even under severe signal degradation, with potential applications in assistive listening technologies.

Published 2026年7月18日