IEEE Trans Neural Syst Rehabil Eng. 2018 Mar;26(3):629-636. doi: 10.1109/TNSRE.2018.2805338.
An electrolarynx (EL) is one of the most popular voice rehabilitation technologies used after laryngectomy. However, most ELs generate monotonic EL speech, which has been shown to create a particular deficit in speech intelligibility, especially for Chinese Mandarin (Mandarin). Mandarin is a tonal language that makes lexical distinctions using variations in tone. Our purpose is to design an EL that can produce the four Mandarin tones, and to evaluate its performance. We designed a fundamental frequency (F0) control method for Mandarin EL speech and manufactured a touch-controlled electrolarynx (T-EL) prototype. Using monosyllables, disyllabic words, and frequently used phrases, we evaluated speech produced with a T-EL, as well as with monotone (M-EL) and variable-frequency modes (P-EL) of a commercially available TruTone EL. A male native Mandarin speaker with laryngectomy volunteered to be the speaker. Results show that the normal speech pitch contours of the four Mandarin tones were most closely matched by the characteristics produced with T-EL. The statistical accuracy of the T-EL's tone and word perception was significantly higher than that of the other EL types. Moreover, the confusion matrix indicates that the listeners could correctly identify the tones of monosyllables and disyllabic words in T-EL speech. Accurate tone judgment can improve the intelligibility of EL speech in Mandarin. The mean opinion score was used to evaluate the listeners' acceptability of EL speech. The scores of the T-EL and M-EL were very close, and the score of the P-EL was significantly lower than that of the other two ELs. However, the results from a single speaker cannot provide sufficient data to conclude which EL has a higher acceptability. The evaluation of multiple EL speakers with different EL types at difference levels of proficiency should be studied in future research.
电子喉（EL）是喉切除术后最常用的语音康复技术之一。然而，大多数 EL 产生单调的 EL 语音，这已被证明会导致语音可懂度特别差，尤其是对汉语普通话（普通话）。普通话是一种声调语言，通过声调的变化来区分词汇。我们的目的是设计一种能够产生四种普通话声调的 EL，并评估其性能。我们设计了一种用于普通话 EL 语音的基频（F0）控制方法，并制造了一个触摸控制电子喉（T-EL）原型。我们使用单音节、双音节词和常用短语评估了 T-EL 产生的语音，以及商业 TruTone EL 的单音（M-EL）和变频模式（P-EL）。一位男性母语为普通话的喉切除术后患者自愿担任演讲者。结果表明，T-EL 产生的特征最接近普通话四个声调的正常语音音高轮廓。T-EL 的声调感知和单词感知的统计准确性明显高于其他 EL 类型。此外，混淆矩阵表明，听众可以正确识别 T-EL 语音中单音节和双音节词的声调。准确的声调判断可以提高普通话 EL 语音的可懂度。平均意见评分用于评估听众对 EL 语音的可接受性。T-EL 和 M-EL 的分数非常接近，而 P-EL 的分数明显低于其他两种 EL。然而，来自单个扬声器的结果不能提供足够的数据来得出哪种 EL 更具可接受性。在未来的研究中，应该研究不同熟练程度的不同 EL 类型的多个 EL 扬声器的评估。
