Some CDs sound great and some don’t: the sound quality of audio programme material is very variable. Expert and naïve listeners are quite good at picking up these differences in sound quality. However, so far there are no metrics that can quantify if a given music track is of good quality or not. This PhD project aims to define and extract quality features from audio signals that enable an automated rating of the acoustic quality therein. The technical aspects of the research project will be underpinned by a substantial study of human factors that determine perceived quality in sound and audio production. The foreseen outcomes are: 1) A framework that sets the relative importance of various objective acoustic measures of signal content in the context of human listening; 2) A digital tool that automatically rates and improves audio quality in a given stream. Applications of the knowledge and technology span from automated adjustment to different reproduction scenarios (eg: radio speech in a car vs. live sound) to archive recovery.