SEMI-BLIND ESTIMATION OF DIRECT-TO-REVERBERANT ENERGY RATIO AND DISTANCE USING RESIDUAL ENERGY TEST STATISTICS

Date

2023-8-29

Author

Gökçe, Ali

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

146
views

63
downloads

Cite This

Direct-to-reverberant ratio (DRR) and the distance of a sound source are attributes of room acoustics and a sound scene recorded in that room, respectively. These different attributes are intricately related: a source at a larger distance is associated with a lower DRR and vice versa. The identification of the source distance and DRR are two problems with use cases in object-based audio systems to enable the reconstruction of scenes recorded by one or more microphone arrays. This thesis proposes a novel approach for a semi-blind estimation of the distance of a sound source and the corresponding direct-to-reverberant ratio in a reverberant room. The proposed method is based on the observations on the correlation between the statistical features of a technique named residual energy test (RENT), and various recording scenarios, i.e. various distances, direct-to-reverberant ratios, room geometries, coupled volumes and the presupposition that systematic changes in the RENT metric can be used to develop a simple approach to estimate DRR and source distance. This proposed approach requires a procedure resembling a calibration step to extract the parameters of a linear regression model. Several recordings from various source-microphone positions in a room are used to extract the statistics that connect the RENT values to distance and direct-to-reverberant ratio value estimates. Once the relation between the RENT statistics and the distance and DRR information is established, the model is then used for estimating the attributes in question. It was observed that for rooms with simpler geometry and volume, the proposed approach affords a mean absolute error (MAE) of 0.5 dB for DRR and 0.07 m for distance. For rooms featuring various materials and coupled volumes, errors can grow with increasing distance, with an MAE of 2.4 dB for DRR and 1.2 m for distance. The performance remains the same for any audio content and is also found to improve when a wider variety of sounds are used in the calibration step.

Subject Keywords

Spherical microphone array, Spherical harmonics, Direct-to-reverberant energy ratio, Room acoustics

URI

https://hdl.handle.net/11511/105379

Collections

Graduate School of Informatics, Thesis