This thesis presents the design, implementation, and evaluation of a solution for managing multiple concurrent audio sources in the GNU/Linux desktop environment. It is intended for everyday computer use, where music, speech, video calls, and notifications are often played simultaneously. When these audio sources are active at the same time, users usually separate them mainly by adjusting their volume. The proposed approach additionally allows users to place the audio sources of individual applications at different virtual positions around the listener.
For such a spatial arrangement, individual audio sources must be processed appropriately before they are mixed into a common headphone output. The PipeWire multimedia server was used for this purpose, as it enables audio streams to be routed through a processing graph. Spatial processing was integrated into this graph as a processing node created with the filter-chain module. For binaural reproduction, existing SOFA filters were used, which process the input signal with impulse responses for the left and right ear. The developed control application uses the PipeWire application programming interface to route application audio streams through the processing node, while the user interface allows individual sources to be arranged in virtual space. The impulse responses are obtained from a SOFA dataset (Spatially Oriented Format for Acoustics), a standardised format for measurements that describe the influence of the head, torso, and pinnae on sound from different directions.
The solution was evaluated in two parts. First, technical measurements assessed operating stability, CPU load, and the spatial processing latency. A user study was then conducted with two practical scenarios in which participants listened to simultaneous audio sources and recorded the information they received. In the first scenario, two speech sources were present; in the second, an additional music source was added to the speech sources. Participants completed both scenarios in two modes: with the audio sources arranged in virtual space and without spatial processing.
The results showed stable operation in the test environment, low CPU load, and no perceived audible artefacts. In the spatial mode, participants correctly understood and recorded more information in both scenarios than in the mode without spatial processing, and the differences between the modes were statistically significant. The System Usability Scale (SUS) results showed that the average usability score was higher in the spatial mode, although the difference was not statistically significant. In contrast, the short version of the User Experience Questionnaire (UEQ-S) showed statistically significant differences in the pragmatic, hedonic, and overall user experience scores.
|
