Room correction: a dynamic, adaptive approach
Lead Research Organisation:
Queen Mary University of London
Department Name: Sch of Electronic Eng & Computer Science
Abstract
This part of the project investigates the use of adaptive filtering applied to sound while the listener is moving in the environment e.g a studio, a room at home. Adaptive filtering is a very active area of research in acoustics as it brings numerous benefits in terms of effectiveness, performance, and better use of computing resources, to name a few. The contribution of this work is that to date, to the best of our knowledge, the work to apply these filters and use them in a dynamic setting has not been done when applying to music a listener listens to in a room and moves around that space freely. The work in the area uses a white noise signal and restricts the movement of the listener to a few fixed locations and so the filtering problem is simpler than when the listener is allowed to move freely and the sound signal is a lot of different frequencies in
them i.e. music in different genres. The existing approach is to limit the listener's movement to a fixed set of locations, and the tests are made using a computer-generated sound with additive white noise [1, 2]. We analyzed the results from different viewpoints. One is the mean squared error (MSE) reported by applying a filter. In this work, we calculate the mean squared error misalignment of the filters. We also applied the DTW (Dynamic Time Warping) algorithm to compare how "similar" the filtered responses are to the original recordings [3]. The DTW algorithm gives a cost indicating the closeness of the match. We then do subjective listening tests where we ask listeners to gauge how close the filtered sound at
a location is to the reference sample. The next part of the project is novel in that we add active noise cancellation to the filtering thus improving the sound.
them i.e. music in different genres. The existing approach is to limit the listener's movement to a fixed set of locations, and the tests are made using a computer-generated sound with additive white noise [1, 2]. We analyzed the results from different viewpoints. One is the mean squared error (MSE) reported by applying a filter. In this work, we calculate the mean squared error misalignment of the filters. We also applied the DTW (Dynamic Time Warping) algorithm to compare how "similar" the filtered responses are to the original recordings [3]. The DTW algorithm gives a cost indicating the closeness of the match. We then do subjective listening tests where we ask listeners to gauge how close the filtered sound at
a location is to the reference sample. The next part of the project is novel in that we add active noise cancellation to the filtering thus improving the sound.
People |
ORCID iD |
Joshua Reiss (Primary Supervisor) | |
Ashish Patel (Student) |
Studentship Projects
Project Reference | Relationship | Related To | Start | End | Student Name |
---|---|---|---|---|---|
EP/V519935/1 | 30/09/2020 | 29/04/2028 | |||
2496680 | Studentship | EP/V519935/1 | 18/01/2021 | 17/01/2028 | Ashish Patel |