Binaural speech intelligibility tests conducted remotely over the Internet compared with tests under controlled laboratory conditions

Abstract

Binaural speech intelligibility tests were carried out with the intention of comparing and assessing the possible similarity of results from tests performed remotely over the Internet, with tests conducted in a controlled laboratory environment. Clean anechoic digital recordings of speech were disturbed with artificial noise or reverberation, interaurally correlated or uncorrelated at the left and right ears. A set of bi-syllable words in Spanish was used as speech material which was diotically delivered through headphones. Internet participants used their own headphones of varying type and audio quality, while participants in the laboratory tests used a single pair of high audio quality circumaural type headphones. Statistical factor analysis was performed on a data set including the test scores, subject age, headphone price (as an indication of sound reproduction quality), and some other factors related to the acoustic stimuli, and to the test conditions. Results from the remote speech intelligibility tests are shown to be consistent with results from tests carried out in the laboratory, with test scores showing small and statistically little significant dependence on headphone price, and laboratory vs. remote test conditions.