Spatial Audio Compression with Adaptive Singular Value Decomposition Using Reconstructed Frames

Namazi M, Elshafiy A, Rose K. Spatial Audio Compression with Adaptive Singular Value Decomposition Using Reconstructed Frames. In: AES Conference: AES 2022 International Audio for Virtual and Augmented Reality Conference. Audio Engineering Society; 2022. Paper 28. Available from: https://aes.org/publications/elibrary-page/?id=21858

@inproceedings{Namazi2022_21858,
  author = {Namazi, Mahmoud and Elshafiy, Ahmed and Rose, Kenneth},
  title = {{Spatial Audio Compression with Adaptive Singular Value Decomposition Using Reconstructed Frames}},
  booktitle = {AES Conference: AES 2022 International Audio for Virtual and Augmented Reality Conference},
  note = {Paper 28},
  year = {2022},
  month = aug,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=21858}
}

TY  - CPAPER
TI  - Spatial Audio Compression with Adaptive Singular Value Decomposition Using Reconstructed Frames
AU  - Namazi, Mahmoud
AU  - Elshafiy, Ahmed
AU  - Rose, Kenneth
T2  - AES Conference: AES 2022 International Audio for Virtual and Augmented Reality Conference
M1  - Paper 28
PY  - 2022
DA  - 2022/08/06
UR  - https://aes.org/publications/elibrary-page/?id=21858
PB  - Audio Engineering Society
LA  - en
AB  - MPEG-H 3D Audio is the current standard for the compression of higher-order ambisonics data. It uses singular value decomposition (SVD) to spatially decorrelate higher-order ambisonics data, followed by the modified discrete cosine transform to exploit temporal decorrelation. Prominent and ambient sound components are then separately encoded (e.g., using the standard core audio codec) and sent to the decoder. Significant improvements in bitrate and audio quality have been gained in earlier work over MPEG-H by applying the SVD operation in the frequency domain rather than the ambisonics domain. In this work, we provide additional compression gains by adaptively calculating and extending the set of SVD basis vectors, at negligible increase in side information cost, using information attained from the previously reconstructed frame. Objective and subjective results provide evidence for higher compression gains when compared to existing methods.
ER  -