Real-time listening experience of feedback-delay network reverberation using geometrical acoustics
Translated title
Real-time lytter oplevelse af feedback-delay netværk reverb og geometrisk akustik
Author
Laursen, Anders Havmøller
Term
4. Term
Education
Publication year
2019
Submitted on
2019-09-11
Pages
35
Abstract
At simulere akustik i realtid er vanskeligt, især når både lydkilde og lytter bevæger sig uafhængigt. Det kræver mange beregninger for at opnå overbevisende resultater. For at reducere beregningsbyrden undersøger dette projekt brugen af radiance-transfer-metoden til at udlede relevante geometriske statistikker for rummet. Disse statistikker bruges derefter til at sætte parametrene i et feedback-delay-netværk (FDN), som er kendt for lav beregningsomkostning og god samlet kvalitet og derfor velegnet til realtidsbrug. Målet er at finde en FDN-orden (modelkompleksitet), der giver en god balance mellem beregningsforbrug og oplevet lydkvalitet. En MURSHA-inspireret test (en sammenlignende lytteprøve) blev brugt til at evaluere forskellige ordener. Testen viste markante forskelle i deltagernes oplevelser for lavere ordener, mens yderligere forøgelse over 8. orden gradvist havde mindre indflydelse på lytteoplevelsen. Der er behov for yderligere tests for at vurdere den samlede lytteoplevelse af systemet sammenlignet med andre systemer.
Simulating acoustics in real time is challenging, especially when the sound source and the listener move independently. This requires many calculations to produce plausible results. To reduce the computational load, this project examines using a radiance-transfer method to derive relevant geometric statistics of the scene. These statistics are then used to set the parameters of a feedback delay network (FDN), a method known for low computational cost and generally good quality, making it suitable for real-time use. The goal is to identify an FDN order (model complexity) that balances computational cost and perceived audio quality. A MURSHA-inspired test (a comparative listening test) was used to evaluate different orders. The test showed clear differences in participants’ experiences for lower orders, while increases beyond the 8th order had gradually less impact on the listening experience. Further testing is needed to assess the overall listening experience of this system compared with other systems.
[This abstract was generated with the help of AI]
Documents