An effective test for the assessment of speech intelligibility in noise with children

Speech intelligibility is degraded under realistic acoustic conditions, as under background noise and reverberation. The assessment of the quality of the listening and speech understanding, both at diagnostic and research purposes, is typically based on standard speech intelligibility tests consisted of single words. However, the available ones have been developed for measurements in quiet and their test-retest reliability is not good enough for accurate assessment of speech recognition in noise or reverberation. An effective tool for internationally comparable assessment of speech intelligibility in noise was developed, that is, the Matrix-type sentence test. This kind of test consisting of 5-word sentences has been developed for about 18 languages. To ensure concentration and easiness in the test administration with the school-aged children, a simplified version of Matrix test has been proposed. This contribution aims at evaluation of the simplified Italian Matrix (ITAMatrix) test. Listening experiments were conducted in two Italian clinics (in Milano and Torino) involving overall 95 children divided in homogenous groups in the 5÷10 years age range. The scope of the measurements was threefold: to state normative data for the assessment of speech intelligibility in noise for children, to establish the reference curve of the psychometric function, and to assess the training effect and test’s reliability in estimating SRT80 (i.e. the speech reception threshold of signal-to-noise ratio, SNR, to yield 80% intelligibility). Across the age groups, a maximum difference of 2 dB SNR in SRT80 and of 2.2%/dB in slope values were found. This strengthens the applicability of simplified ITAMatrix test even with pediatric population of young age (e.g. 5-6 years old). No considerable differences were found across two measurement sites. A high test-retest reliability of 1.1 dB at SRT80 confirms the suitability of this test for accurate measurements in noise.