Development of an auditory-visual matrix sentence test in New Zealand English

• Main Authors: O’Beirne, Greg A., Trounson, Ronald H, McClelland, Amber D, Jamaluddin, Saiful Adli, Maclagan, Margaret A
• Format: Article
• Language: English
• Published: The European Academy of Otology and Neurotology 2015
• Subjects: RF Otorhinolaryngology
• Online Access: http://irep.iium.edu.my/66050/1/O%27Beirne%20et%20al.%20-%202015%20-%20Development%20of%20an%20auditory-visual%20matrix%20sentence%20test%20in%20New%20Zealand%20English.pdf; http://irep.iium.edu.my/66050/; http://www.advancedotology.org/sayilar/86/buyuk/IAO-EFAS-2.pdf

Matrix sentence tests (Hagerman, 1982) consist of syntactically fixed but semantically unpredictable sentences, each composed of 5 words (name, verb, number, adjective, object). Test sentences are generated by choosing one of ten alternatives for each word to form up to 100,000 unique sentences (e.g. “Amy bought six dark hats”). Rather than recording these sentences individually, the sentences are synthesised from 400 recorded audio fragments that preserve coarticulations and provide a natural prosody for the synthesised sentence (Wagener, 1999). Matrix tests produced in this way have been developed in numerous European languages. We describe here the development of an adaptive speech-in-noise matrix sentence test in New Zealand English. The matrix words were modified from the British English matrix (Hewitt, 2007) to be appropriate for both New Zealand and Australian English. However, we have made the important step of getting this type of matrix test to work in an auditory-visual mode for the first time, using recorded fragments of high-definition video. The key to the success of the auditory-visual test mode was the selection of edit points appropriate for both speech and facial expression, and the maintenance of the speaker’s head in a constant position throughout the recording. Normalisation was achieved by generating fragment-specific intelligibility functions for the auditory-alone test mode in the closed-set presentation format, using both speech noise and multi-talker babble. This test shows great clinical promise, as auditory-visual tests are more representative of real-world communication than auditory-alone tests, and allow the assessment of a client’s auditory-visual integration abilities.