van As Corina J, Koopmans-van Beinum Florien J, Pols Louis C W, Hilgers Frans J M
Department of Otolaryngology-Head and Neck Surgery, Netherlands Cancer Institute/Antoni van Leeuwenhoek Hospital, Institute of Phonetic Sciences, University of Amsterdam.
J Speech Lang Hear Res. 2003 Aug;46(4):947-59. doi: 10.1044/1092-4388(2003/074).
The present study was conducted to investigate voice quality in tracheoesophageal speech by means of perceptual evaluations and to develop a clinically useful subset of perceptual scales sufficient for these perceptual evaluations. The perceptual ratings were obtained from both naive and trained raters (speech-language pathologists [SLPs]) after listening to a read-aloud text. The perceptual evaluations were performed by means of 19 semantic bipolar 7-point scales for the naive raters and 20 semantic bipolar 7-point scales for the trained raters. The trained raters were also asked to judge the overall voice quality as good, reasonable, or poor. Both naive listeners and trained SLPs were able to perform reliable perceptual judgments. Naive raters judged the tracheoesophageal voice as more deviant than the trained raters did. Naive raters made judgments based on 2 underlying perceptual dimensions (voice quality and pitch), whereas the trained raters made judgments based on 4 underlying perceptual dimensions (voice quality, tonicity, pitch, and tempo). These perceptual dimensions were further subdivided into a subset of 4 perceptual scales for the naive raters and a subset of 8 perceptual scales for the trained raters. This appeared to provide a sufficient coverage of the underlying perceptual dimensions used by the listeners.
本研究旨在通过感知评估来调查气管食管语音的语音质量，并开发一套对这些感知评估足够有用的临床感知量表子集。在聆听一段朗读文本后，从未经训练的评分者和经过训练的评分者（言语语言病理学家[SLP]）那里获得感知评分。感知评估通过19个语义双极7点量表对未经训练的评分者进行，以及20个语义双极7点量表对经过训练的评分者进行。还要求经过训练的评分者将整体语音质量判断为好、一般或差。未经训练的听众和经过训练的SLP都能够进行可靠的感知判断。未经训练的评分者判断气管食管语音比经过训练的评分者认为的更偏离正常。未经训练的评分者基于2个潜在的感知维度（语音质量和音高）进行判断，而经过训练的评分者基于4个潜在的感知维度（语音质量、张力、音高和节奏）进行判断。这些感知维度进一步细分为未经训练的评分者的4个感知量表子集和经过训练的评分者的8个感知量表子集。这似乎为听众使用的潜在感知维度提供了足够的覆盖范围。
