在训练感知频谱移位噪声编码语音时，比较实时语音和录音语音。

Comparing live to recorded speech in training the perception of spectrally shifted noise-vocoded speech.

机构信息

Speech Hearing and Phonetic Sciences, University College London, Chandler House, 2 Wakefield Street, London WC1N 1PF, United Kingdom.

出版信息

J Acoust Soc Am. 2012 Oct;132(4):EL336-42. doi: 10.1121/1.4754432.

DOI:10.1121/1.4754432

PMID:23039574

Abstract

Two experimental groups were trained for 2 h with live or recorded speech that was noise-vocoded and spectrally shifted and was from the same text and talker. These two groups showed equivalent improvements in performance for vocoded and shifted sentences, and the group trained with recorded speech showed consistently greater improvements than untrained controls. Another group trained with unshifted noise-vocoded speech improved no more than untrained controls. Computer-based training thus appears at least as effective as labor-intensive live-voice training for improving the perception of spectrally shifted noise-vocoded speech, and by implication, for training of users of cochlear implants.

摘要

两个实验组接受了 2 小时的训练，使用来自同一文本和说话者的、经噪声编码和频谱移位的实时或录音语音。这两个组在处理经噪声编码和频谱移位的句子时表现出相当的提高，而使用录音语音进行训练的组比未经训练的对照组表现出持续更大的提高。另一组使用未经移位的噪声编码语音进行训练，提高程度不超过未经训练的对照组。因此，基于计算机的训练至少与劳动密集型的实时语音训练一样有效，可以改善对频谱移位噪声编码语音的感知，并且可以为人工耳蜗使用者的训练提供启示。