Evaluation of spatial audio quality in non-standardized test environments

点击翻译按钮获取中文摘要

Audio quality perception is already subject of different evaluation for a long time. Especially developments in the last few years in the domain of multichannel audio theoretically enable a realistic audio reproduction, but its quality for practical applications still needs to be evaluated. Despite extensive research on spatial audio quality there are still aspects that are not yet fully understood, for example the composition of the quality parameters. Quality evaluations are generally done in standardized test environments, hereinafter referred to as laboratory. Furthermore, mainly standardized evaluation methods are used, which usually measure the basic audio quality compared to a reference signal. The unidimensional, quantitative results of these methods are only comparable and repeatable in standardized test environments. They neither take into account the multidimensionality of audio quality nor the actual context of use. Furthermore, individual quality scales of subjects are not considered and quality ratings are averaged over all subjects. Another critical aspect is that the reference signal is not always known or defined. Recent research that tries to understand the multidimensionality of audio quality perception uses descriptive evaluation methods. These methods allow an individual quality rating often without a reference and participants own descriptions. Different studies already showed that timbre and spaciousness are two very important aspects of spatial audio quality. However, there is still some disagreement on the type and weighting of the quality elements and their influence on the overall quality. Except some evaluations in concert halls, the context of use it not evaluated in existing evaluations so far. Though, the subjective audio quality very much depends on the room acoustics and cannot be transferred simply to other rooms. In this project valid and repeatable evaluation and analysis methods will be developed to better understand human quality perception and to rate the perceived spatial audio quality. The goal of the project is to analyze the dimensions of audio quality perception as well as to evaluate the influence of the test environment on quality perception. The results will show the weighting and influence of the different quality parameters dependent on the context of use. They can be used directly for the development of objective quality evaluation methods and to improve coding algorithms.