Audio quality in everyday listening situations

Abstract

The quality of digital audio which uses a form of lossy data compression is in the focus of several studies. The experimental focus of these studies includes the fundamental difference between objective and subjective measurement methods (Pocta & Beerends, 2015), test standards (Breebaart, 2015), or signal processing (Khaldi et al., 2013; Jung et al., 2016). But there are only few studies concerning the impact of perceived quality differences in everyday listening situations (ELS). Short excerpts of recent popular music recordings were encoded with different bit rates based on codecs that are commonly used by popular music streaming services. While the encoded files were presented via standard Samsung in-ear headphones, the everyday listening situation was simulated by using binaural recordings of a shopping mall soundscape that was simultaneously presented via Stax electrostatic headphones. In a control condition no soundscape was used. The stimuli were rated by 30 participants in a MUSHRA listening test (ITU-R BS.1534). A repeated measures ANOVA showed that there are significant differences for both conditions concerning the different stimuli (ELS: F(10,20) = 25.237, p = .0001, μ² = .927; no ELS: F(10,20) = 51.073, p = .0001, μ² = .962), but the effect sizes are consistently smaller in the condition with simulated everyday listening situation. The difference between both conditions gets smaller, the lower the bit rates are. A follow-up study additionally includes binaural audio that is currently used by radio stations.