DOA-Informed Self-Supervised Learning Method for Sound Source Enhancement

Gao S, Wu X, Qu T. DOA-Informed Self-Supervised Learning Method for Sound Source Enhancement. In: Convention Paper. Audio Engineering Society; 2024. Paper 10683. Available from: https://aes.org/publications/elibrary-page/?id=22496

@inproceedings{Gao2024_22496,
  author = {Gao, Shan and Wu, Xihong and Qu, Tianshu},
  title = {{DOA-Informed Self-Supervised Learning Method for Sound Source Enhancement}},
  note = {Paper 10683},
  year = {2024},
  month = jun,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=22496}
}

TY  - CPAPER
TI  - DOA-Informed Self-Supervised Learning Method for Sound Source Enhancement
AU  - Gao, Shan
AU  - Wu, Xihong
AU  - Qu, Tianshu
M1  - Paper 10683
PY  - 2024
DA  - 2024/06/06
UR  - https://aes.org/publications/elibrary-page/?id=22496
PB  - Audio Engineering Society
LA  - en
AB  - The multiple-channel[1] sound source enhancement methods have made a great progress in recent years, espe-cially when combined with the learning-based algorithms. However, the performance of these techniques is lim-ited by the completeness of the training dataset, which may degrade in mismatched environments. In this paper, we propose a reconstruction Model based Self-supervised Learning (RMSL) method for sound source enhance-ment. A reconstruction module is used to integrate the estimated target signal and noise components to regenerate the multi-channel mixed signals, and it is connected with a separating model to form a closed loop.In this case, the optimization of the separation model can be achieved by continuously iterating the separation-reconstruction process. We use the separation error, the reconstruction error, and the signal-noise independence error as loss functions in the self-supervised learning process. This method is applied to the state-of-the-art sound source separation model (ADL-MVDR) and evaluated under different scenarios. Experimental results demonstrate that the proposed method can improve the performance of ADL-MVDR algorithm under different number of sound sources, bringing about 0.5 dB to 1 dB Si-SNR gain, while maintaining good clarity and intelligibility in practical application.
ER  -