Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
portfolio
publications
A study on deep neural network based speech enhancement by combining outlier-robust loss
Published in The Journal of the Acoustical Society of Korea, 44(1), 2025
Replacing the frequency-domain MSE term in mask-based speech enhancement with the outlier-robust Huber and Berhu losses improves SDR, PESQ and STOI together on VCTK, beating the conventional SI-SNR + MSE objective.
Recommended citation: Bongsu Jung and Wooil Kim. (2025). "A study on deep neural network based speech enhancement by combining outlier-robust loss." The Journal of the Acoustical Society of Korea, 44(1), pp. 66–73. doi:10.7776/ASK.2025.44.1.066
Download Paper | Download Bibtex
DLF-EEND: Dynamic Layer Fusion for End-to-End Speaker Diarization
Published in Interspeech 2025, Rotterdam, The Netherlands, 2025
An auxiliary training branch dynamically routes and fuses multi-resolution Transformer representations at each time step, cutting DER by 59.18% on simulated data and improving CALLHOME by 14.26% over SA-EEND – at no inference cost, since only the main branch is used at test time.
Recommended citation: Bongsu Jung and Wooil Kim. (2025). "DLF-EEND: Dynamic Layer Fusion for End-to-End Speaker Diarization." Interspeech 2025, pp. 1688–1692. doi:10.21437/Interspeech.2025-1059
Download Paper | Download Bibtex
Hierarchical Permutation Consistency Learning for Self-Conditioned End-to-End Speaker Diarization
Published in Interspeech 2026, Sydney, Australia, 2026
Layer-wise PIT picks each layer’s speaker permutation independently, so self-conditioned EEND feeds permutation noise back into its own conditioning path. Group-Wise PIT and Continuous Influence PIT constrain permutations across the hierarchy instead, stabilising training and improving DER on CALLHOME for 2- and 3-speaker mixtures.
Recommended citation: Bongsu Jung and Wooil Kim. (2026). "Hierarchical Permutation Consistency Learning for Self-Conditioned End-to-End Speaker Diarization." Interspeech 2026, pp. 5307–5311. doi:10.21437/Interspeech.2026-1198
Download Paper | Download Bibtex
Clustering-Free End-to-End Spoof Diarization with Encoder-Decoder Attractors
Published in NeurIPS 2026, Sydney, Australia (to appear), 2026
Accepted to the NeurIPS 2026 Main Track as a poster.
Recommended citation: Bongsu Jung, Donghee Kim, and Wooil Kim. (2026). "Clustering-Free End-to-End Spoof Diarization with Encoder-Decoder Attractors." Advances in Neural Information Processing Systems (NeurIPS). To appear.
Download Bibtex