On listeners preferences in subjective assessment tests of multichannel audio on mobile phones

Abstract

ITU standard listening tests like the ITU-R BS.1534-3 (MUSHRA) emphasize on the use of expert listeners especially for rating global attributes like Basic Audio Quality. The ratings of expert listeners are more reliable but naive listeners are often used for such tests, especially when attributes like Quality of Experience are being measured. Moreover, the results from naive listeners are actually considered more reflective of average consumer preference. This paper presents further statistical analysis of the results of two such previously conducted tests for assessing multichannel audio versus stereo and mono on mobile phones using headphones. The procedure and materials used in the tests were compliant with the MUSHRA requirements and the test material consisted of audio excerpts with 6 or 8 channels. The multichannel excerpts were object coded and served as the reference which in turn was further down-mixed to different versions of stereo and mono to serve as the medium and low anchors. The media player used had custom HRTF's and rendered the multichannel excepts for a binaural playback on headphones. The analysis involved measure of rank correlations on the results in order to study the listener ratings in more detail. The results showed that for the ratings in both tests, the preference for multichannel, stereo and mono had little correlation with age and gender but a regression model for predicting ratings of one test from the ratings of the other was possible.