Room Estimation
Translated title
Rum Estimering
Authors
Jøns, Christian Suhr ; Dominguez Simon, Jose Ignacio
Term
4. term
Education
Publication year
2016
Submitted on
2016-06-02
Pages
151
Abstract
Denne afhandling undersøger, om man kan estimere et rums geometri og hvor meget væggene absorberer lyd, baseret på en akustisk måling. Vi ser på, hvordan lydfeltet i et rum kan beskrives og genskabes ud fra målinger. Lydfeltet består af den direkte lyd fra kilden og refleksioner fra overflader. Refleksionerne modelleres som virtuelle lydkilder, og hvis deres retning og afstand bestemmes præcist, kan lydfeltet estimeres godt. I simulerede målinger anvendes en cirkulær mikrofonkreds (mikrofoner arrangeret i en ring) og array-signalbehandling til at estimere ankomstretning (Direction of Arrival, DOA) for de virtuelle kilder. En stokastisk maksimum-likelihood-estimator bruges til DOA-estimeringen. Ved at plotte fordelingen af lydfeltets effekt kan rummets geometri udledes. For at vurdere væggenes absorption sammenlignes frekvensspektret fra førsteordens virtuelle kilder (de tidligste refleksioner) med spektret af den direkte lyd, med lovende resultater.
This thesis investigates whether a room's geometry and wall absorption can be estimated from an acoustic measurement. It examines how the sound field in a room can be described and reconstructed from measurements. The sound field is made up of the direct sound from the source and reflections from surfaces. Reflections are modeled as virtual sound sources, and if their directions and distances are determined accurately, the sound field can be well estimated. In simulated measurements, a circular microphone array (microphones arranged in a ring) and array signal processing are used to estimate the direction of arrival (DOA) of the virtual sources. A stochastic maximum likelihood estimator is applied for DOA estimation. By plotting the power distribution of the reconstructed sound field, the room's geometry can be inferred. To assess wall absorption, the frequency spectrum of first-order virtual sources (the earliest reflections) is compared to the spectrum of the direct sound, with promising results.
[This abstract was generated with the help of AI]
Keywords
Documents