This thesis explores how the study of realistic mixes can expand current knowledge about multitrack music mixing. An essential component of music production, mixing remains an esoteric matter with few established best practices. Research on the topic is challenged by a lack of suitable datasets, and consists primarily of controlled studies focusing on a single type of signal processing. However, considering one of these processes in isolation neglects the multidimensional nature of mixing. For this reason, this work presents an analysis and evaluation of real-life mixes, demonstrating that it is a viable and even necessary approach to learn more about how mixes are created and perceived. Addressing the need for appropriate data, a database of 600 multitrack audio recordings is introduced, and mixes are produced by skilled engineers for a selection of songs. This corpus is subjectively evaluated by 33 expert listeners, using a new framework tailored to the requirements of comparison of musical signal processing. By studying the relationship between these assessments and objective audio features, previous results are confirmed or revised, new rules are unearthed, and descriptive terms can be defined. In particular, it is shown that examples of inadequate processing, combined with subjective evaluation, are essential in revealing the impact of mix processes on perception. As a case study, the percept `reverberation amount' is ex-pressed as a function of two objective measures, and a range of acceptable values can be delineated. To establish the generality of these findings, the experiments are repeated with an expanded set of 180 mixes, assessed by 150 subjects with varying levels of experience from seven different locations in five countries. This largely confirms initial findings, showing few distinguishable trends between groups. Increasing experience of the listener results in a larger proportion of critical and specific statements, and agreement with other experts.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:765949 |
| Date | January 2017 |
| Creators | De Man, Brecht |
| Publisher | Queen Mary, University of London |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://qmro.qmul.ac.uk/xmlui/handle/123456789/25814 |
Page generated in 0.002 seconds