從近似到驗證：重新檢視頻率匹配的人工耳蝸聲學模型

// From Approximation to Validation: Rethinking Frequency-Matched Acoustic Models of Cochlear Implants

44名單側失聰CI使用者自調vocoder（語音模擬器），假設須納入聽者特定頻率失配｜結果自選模型被偏好、聽感相似度高（平均6.11/7）且語音分數與CI一致｜頻率對齊模型評分低且高估語音表現，顯示需個人化頻率失配以有效模擬CI，具應用價值。

// This study tested whether vocoder-based acoustic models of cochlear implants capture both perceived sound quality and speech performance, hypothesizing that listener-specific frequency mismatch is required. Forty-four single-sided-deaf CI users created self-selected vocoders by adjusting low/high cutoff and channel interaction, then rated those and several frequency-matched tone/noise vocoders and completed speech tests. Nearly all users preferred models with listener-specific frequency shifts and minimal channel interaction; self-selected models were rated much more similar (mean 6.11/7) and matched CI speech scores, while frequency-matched models were rated poorly and significantly overestimated speech performance. Therefore, frequency-matched vocoders do not validly model CI sound or performance, whereas listener-specific mismatch yields substantially better perceptual and speech matches, though some factors remain unaccounted for.

Published 2026年9月16日