2021 Fiscal Year Final Research Report
Cross-modal signal estimation by coupled dictionary learning and its aaplication to non-contact sensing
Project/Area Number |
19K04429
|
Research Category |
Grant-in-Aid for Scientific Research (C)
|
Allocation Type | Multi-year Fund |
Section | 一般 |
Review Section |
Basic Section 21030:Measurement engineering-related
|
Research Institution | Chiba Institute of Technology |
Principal Investigator |
|
Project Period (FY) |
2019-04-01 – 2022-03-31
|
Keywords | 教師なし学習 / 信号分離 / 光電脈波 / 生体信号 / 信号推定 / スパース信号処理 / 辞書学習 |
Outline of Final Research Achievements |
In this study, applications of convolutional dictionary learning for non-contact sensing of human activities. The convolutional dictionary learning decomposes a signal into a set of component signals under sparsity prior. In this study, the dictionary learning is simultaneously applied to the contact sensor and the non-contact sensor. By using the proposed combined dictionary learning, the signal from the contact sensor signal is estimated from the non-contact sensor by using the combined dictionary that is prior trained. The proposed group sparse dictionary learning can obtain the pair of the dictionary that are correlated. By using this the signal that is highly correlated to the contact-sensor can be estimated from the non-contact sensor. We apply the proposed method to the remote PPG(photo-plethysmography) and the micro-wave Doppler sensor for detection of breathing. By these application, we demonstrate the estimation results can be improved by using two-different sensors.
|
Free Research Field |
信号処理
|
Academic Significance and Societal Importance of the Research Achievements |
本研究で提案した結合畳み込み型辞書学習および部分空間スパース辞書学習により,複数の信号を同時に取り扱うことで,信号間に共通した特徴を取り出すことが可能になった.この特長により,従来までは,センサ信号から,目的とする生体信号などの特徴を示す信号を取り出すために経験的な後処理が必要であったが,ノイズの少ない接触型センサの信号を利用する事前学習により,自動的に精度よく信号が分離できるようになった.これを既存の脈拍,呼吸などの遠隔センシングに利用することで,既存のセンサからの推定精度を向上させることができる.
|