컨텐츠 시작
학술대회/행사
초록검색
| 제출번호(No.) | 0223 |
|---|---|
| 분류(Section) | Invited Lecture |
| 분과(Session) | (AM) Applied Mathematics(including AI, Data Science) (AM) |
| 발표시간(Time) | 20th-O-15:50 -- 16:30 |
| 영문제목 (Title(Eng.)) |
Data science applied to wearables for personalized digital healthcare |
| 저자(Author(s)) |
Dae Wook Kim1, Minki P. Lee2, Daniel B. Forger2 Sogang University1, University of Michigan2 |
| 초록본문(Abstract) | Currently, millions of individuals utilize wearable devices like the Apple Watch to track their physical activity, heart rate, and other physiological signals, resulting in a vast amount of wearable data. This influx of data provides a unique opportunity for digital medicine to advance precision healthcare. However, the inherent noise in this data poses a challenge, rendering it seemingly unusable without the development of new mathematical techniques for signal extraction. In this talk, I will present several techniques we have devised for analyzing this noisy time-series data. These include the Kalman filter-based data assimilation method, which serves as a novel state space estimation technique capable of estimating phases of circadian rhythms. Additionally, I will introduce a Kalman filter-assisted autoencoder for anomaly detection in time-series data, along with feature engineering methods grounded in persistent homology and mathematical modeling. These techniques offer practical applications such as sleep quality assessment, early detection of physiological changes linked to fever and daily mood prediction. |
| 분류기호 (MSC number(s)) |
37N25, 68T07, 92B25, 92C30, 92-08, 92-10 |
| 키워드(Keyword(s)) | Data assimilation, Kalman filter, nonlinear estimation, neural network,mathematical models, wearables, circadian rhythms |
| 강연 형태 (Language of Session (Talk)) |
Korean |