Development of the computerized Cantonese Disyllabic Lexical Tone Identification Test in Noise (CANDILET-N)

Abstract

CANDILET-N is a closed-set speech-recognition test to assess lexical tone recognition in noise for Cantonese speakers. The test consists of 60 test items in a four-alternative forced-choice test paradigm, with male and female speaker versions. Each test plate consists of two disyllabic-word lexical tone minimal pair test items and their respective phonemic distracters. Psychometric performance SNR (signal-to-noise ratio) functions from 26 normal hearing adult subjects were fitted to a logistic regression model to obtain the SNR for 50% correct score (SNR-50%) for individual test items. The mean SNR-50% of the 60 test items for the female and male test sets were – 10.3 dB and – 11.1 dB, respectively. The SNR-50% varies across test items and their SNR need to be adjusted to improve the homogeneity among them. Copyright © 2009 John Wiley & Sons, Ltd.

Keywords:

• Cantonese
• lexical tone
• speech recognition
• signal-to-noise ratio for 50% correct score (SNR-50%)
• test development