Creation of dynamic sound zones with adaptive filters
Authors
Jensen, Mads Hulegaard ; Damgaard, Thomas
Term
4. term
Education
Publication year
2020
Submitted on
2020-06-04
Pages
67
Abstract
I dette speciale udvikles adaptive FIR-filtre i frekvensdomænet, der kan forme dynamiske lydzoner i små, afgrænsede områder ved at filtrere afspilningssignalerne, før de gengives. På baggrund af de filtrerede signaler foreslås to dynamiske modeller, der forudsiger det akustiske tryk i punkter, som kan ændre sig over tid. Modellerne bygger på to standardantagelser: fjernfelt (punkterne er relativt langt fra kilderne) og efterklangsfri forhold (ingen refleksioner). Filtrene beregnes ved iterativt at løse et optimeringsproblem, så forskellen mellem det modellerede, tidsvarierende tryk og et ønsket, tidsvarierende reference-tryk minimeres. Herved kan filtrene løbende tilpasse sig de dynamiske lydzoner. De to modeller adskiller sig ved, om de medtager trykbidrag fra andre afspilningssignaler eller behandler hvert signal uafhængigt af de øvrige. Begge metoder gav samme reproduktionsfejl og næsten samme array effort (den samlede indsats i arrayet). De foreslåede adaptive filtre er desuden hukommelseseffektive, og opdateringen kan i høj grad paralleliseres.
This thesis develops adaptive finite impulse response (FIR) filters that operate in the frequency domain to form dynamic sound zones within small, confined regions by filtering the playback signals before reproduction. Based on these filtered signals, two dynamic models are proposed to predict the acoustic pressure at spatial points that may change over time. The models rely on two simplifying assumptions: far-field (points are relatively far from the sources) and anechoic conditions (no reflections). The filters are obtained by iteratively solving an optimization problem that minimizes the difference between the modeled, time-varying pressure and a desired, time-varying reference pressure. This makes the filters adaptable as the sound zones change. The two models differ in whether they include pressure contributions from other playback signals or treat each signal independently. Both methods achieved the same reproduction error and almost the same array effort. The proposed adaptive filters require little memory, and the update procedure can be highly parallelized.
[This abstract was generated with the help of AI]
Keywords
Documents