Development of pattern recognition algorithm for ultra low frequency multivariate time-series data considering dimensional correlation

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K11938
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61010:Perceptual information processing-related
• Research Institution: Chuo University (2022-2023); Yokohama City University (2021)
• Principal Investigator: OKUSA KOSUKE 中央大学, 理工学部, 准教授 (30636907)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: マイクロ波センサ / 数理モデリング / バイタル検知 / 非接触 / 心拍 / RRI / 血圧

Outline of Final Research Achievements

When performing pattern recognition using multidimensional time-series data, it is crucial to consider the frequency characteristics and the correlations between dimensions. However, when the observed sequences, such as human motions, are short-term and ultra-low frequency signals, these features are not easily extractable, making it exceedingly difficult to analyze them while taking these aspects into account. This study aims to develop an analysis method that can stably infer features under the constraint of the observed sequences being multidimensional ultra-low frequency stream data. Specifically, the research focused on vital detection such as heart rate and blood pressure, conducting experiments and validations. By incorporating mathematical models using information obtained from multidimensional sensor data with one or multiple microwave sensors, the study aims to improve measurement accuracy.

Free Research Field

センシングデータ解析

Academic Significance and Societal Importance of the Research Achievements

本研究での研究成果は，超高齢化社会を迎える日本において非常に重要な役割を果たす．特に本研究での実用例で実施したバイタルセンシングは，独居老人の見守りシステムや，普段からの健康管理に重要な役割を果たすと考えられる．非接触型センサによる高精度バイタルセンシングを実現したことにより，センサを身につける煩わしさのない，普段の生活での健康状態の把握が実現できるものと考えられる．