Psychoacoustic characterisation of linguistically balanced, Danish sentences for speech-in-noise experiments

Abstract

Objective: To psychoacoustically characterise a corpus of Danish sentences that have been developed for speech-in-noise experiments and to assess the influence of speech intelligibility scoring methods and linguistic syntax on the resulting psychometric functions. Design: Psychometric functions are estimated for each sentence in stationary speech-shaped noise using the method of constant stimuli and five signal-to-noise ratios from −10 to 0 dB. Study sample: Sentences from one of the four talkers in the corpus were presented pseudo-randomly to 60 normal hearing participants so that each sentence was heard by 20 different participants. Results: The inflection point of the psychometric function was affected by both the scoring rule applied and by the number of scoring units per sentence, but not by the linguistic syntax. The slope of the psychometric function was significantly affected by only the number of scoring units. The mean psychometric function based on one of the scoring methods revealed a speech reception threshold ((Formula presented.)) of −4.8 dB and a slope of 11.7%/dB. Conclusions: The psychometric functions of the sentences are modifiable depending on the scoring method applied. Moreover, the linguistic variation of the syntax present in the corpus does not lead to systematic changes in the psychometric functions.