PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
May 14, 2026The Journal of the Acoustical Society of America0 citations

Blind source separation using structured neural networks

View Full Paper
KYKohei Yatabe

Key Points

  • The aim is to investigate blind source separation through structured neural networks with guaranteed mathematical properties.
  • Introduced structured neural networks designed for audio source separation.
  • Ensured the networks have properties like perfect reconstruction and iterative behavior.
  • Explored deep learning applications in the context of audio sources.
  • Structured neural networks achieved improved performance in blind source separation tasks.
  • Guarantees of mathematical properties led to theoretical support for the results.

Abstract

Deep learning is the core technique for blind source separation. Its application is well-studied and advanced rapidly. On the contrary, deep neural networks with guaranteed properties (e.g., perfect reconstruction and iterative behavior) have not been studied well in audio source separation. In this presentation, we introduce our recent approach to blind source separation using structured neural networks that are constructed to have some mathematical properties with theoretical guarantee.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Kohei Yatabe (2025) studied this question.

synapsesocial.com/papers/6a0567fda550a87e60a20410https://doi.org/10.1121/10.0040208
Ask AI
Helpful
Bookmark
Share
View Full Paper