Extreme Signal Processing

2022 Fiscal Year Final Research Report

• Project/Area Number: 20H02145
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Review Section: Basic Section 21020:Communication and network engineering-related
• Research Institution: Osaka University (2022); Tokyo University of Agriculture and Technology (2020-2021)
• Principal Investigator: Tanaka Yuichi 大阪大学, 大学院工学研究科, 教授 (10547029)
• Co-Investigator(Kenkyū-buntansha): 田中 聡久 東京農工大学, 工学(系)研究科(研究院), 教授 (70360584) ; 石田 寛 東京農工大学, 工学(系)研究科(研究院), 教授 (80293041) ; 小野 峻佑 東京工業大学, 情報理工学院, 准教授 (60752269)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 信号処理 / 深層学習

Outline of Final Research Achievements

We conducted research aimed at creating a data analysis technology for the restoration of "extreme signals," i.e., sensor data obtained from environments with very low signal-to-noise ratios and approaching a "once-in-a-lifetime" situation, as well as knowledge discovery and information extraction from extreme signals. We focused on graph signal processing and deep unrolling, and theoretically investigated the graph sampling theorem. We demonstrated the superior performance of methods using deep unrolling for sensor data restoration and interpolation/noise removal of data on time-varying graphs compared to conventional methods.

Free Research Field

信号情報処理

Academic Significance and Societal Importance of the Research Achievements

信号処理技術と深層学習技術を適切に組み合わせ，利用することで，様々な劣化状態のデータ（信号）に対して優れた修復手法が実現できることを明らかにした．特に，サンプリング定理や深層展開のグラフ上データへの拡張に関して成果を挙げた．本研究から得られた成果は，今後必須となるグリーンなデータ解析技術の嚆矢となる技術である．