Real-time listening experience of feedback-delay network reverberation using geometrical acoustics

Laursen, Anders Havmøller

4. Term

Sound and Music Computing

2019

2019-09-11

35 pages

Akustisk modelering er kompleks og møder ofte besværligheder når det skal køre i real-tid, når både kilde og lytter dynamisk bevæger sig uafhængigt af hinanden. Store mængder af beregninger er påkrævet ved run-time for at skabe troværdige resultater. For at imødekomme mængden af beregninger som er påkrævede, forsøger dette studie at benytte radiance transfer for at beregne relevant geometrisk statistik, som så kan benyttes til at instille et feedback-delay netværk, ved brug af denne statistik. Feedback-delay netværk er kendte for deres hurtige beregninger og lyd-kvalitet, hvilket gør dem meget anvendelige i real-tids applikationer. Projektet forsøger at finde den orden af feedback-delay netværk som er både effektivt, og har subjektivt god lydkvalitet. En MURSHA-inspireret test viste at signifikante forskelle i oplevet lydkvalitet mellem lavere systemer af lavere orden. Det blev også vist, at det gradvist har miondre indflydelse på oplevelsen af kvalitet, når højere orden systemer sammenlignes, over 8-ordens systemer. Flere eksperimenter er dog påkrævet for at kunne sige noget om systemets generelle kvalitet.

Acoustical modelling is complex and often face difficulties when having to run in real time, when both source and listener are moving dynamically and independently. Large amounts of calculations are required at run-time in order to produce plausible results. In order to mitigate the amount of calculations required at run-time, this paper investigates the use of radiance-transfer method to calculate relevant geometrical statistics, which can then be used to set the parameters of a feedback-delay network according to said statistics. Feedback delay networks are known for their low computational cost, and generally good overall quality and are thus popular in real-time applications. This project seeks to find an order of feedback-delay network that sits at a good middle ground between computational cost, and subjective quality experience. A MURSHA inspired test revealed that significant differences in participant experiences are present when comparing lower order systems. Testing also revealed that it becomes gradually less impactful to the listening experience, when comparing higher order systems, beyond 8th order. Further testing is required for the overall listening experience of the system itself compared to other systems.

Download
View record in AAU Student Projects