Remote Blood Pressure Estimation via the Spatiotemporal Mapping of Facial Videos.

Blood pressure (BP) monitoring is vital in daily healthcare, especially for cardiovascular diseases. However, BP values are mainly acquired through a contact-sensing method, which is inconvenient and unfriendly for BP monitoring. This paper proposes an efficient end-to-end network for estimating BP values from a facial video to achieve remote BP estimation in daily life. The network first derives a spatiotemporal map of a facial video. Then, it regresses the BP ranges with a designed blood pressure classifier and simultaneously calculates the specific value with a blood pressure calculator in each BP range based on the spatiotemporal map. In addition, an innovative oversampling training strategy was developed to handle the problem of unbalanced data distribution. Finally, we trained the proposed blood pressure estimation network on a private dataset, MPM-BP, and tested it on a popular public dataset, MMSE-HR. As a result, the proposed network achieved a mean absolute error (MAE) and root mean square error (RMSE) of 12.35 mmHg and 16.55 mmHg on systolic BP estimations, and those for diastolic BP were 9.54 mmHg and 12.22 mmHg, which were better than the values obtained in recent works. It can be concluded that the proposed method has excellent potential for camera-based BP monitoring in the indoor scenarios in the real world.

Privacy-preserving state estimation with unreliable channels.

A co-design problem of privacy-preserving encoding and filtering is concerned in this paper. A remote legal user estimates states of a dynamic system by the received innovation messages. However, the communication channel is neither reliable nor secure. The messages may be missing during the transmission and have a risk of being intercepted by an eavesdropper. Therefore, in this work, an encoding scheme together with an MMSE estimation algorithm are designed for preserving information privacy and meanwhile guaranteeing estimation performance. We introduce a novel encoding approach using a weighted innovation with a public key to achieve privacy security as well as low computational expense. A filtering algorithm is designed under such an encoding mechanism. And the performance of the filter and the encoding method is analyzed. The feasibility of the encoding approach and the filtering algorithm is illustrated through numerical examples of an unstable system and a stable ballistic roll rate estimation system.

A unified model for high resolution mapping of global lake (>1 ha) clarity using Landsat imagery data.

Lake clarity, usually measured by Secchi disc depth (SDD), is a reliable proxy of lakes trophic status due to its close link with total suspended matter, chlorophyll-a, and nutrients. Trained with in-situ measured SDD and match-up Landsat images, we established various regression models to estimate SDD for global lakes. We selected a unified model which demonstrated good spatiotemporal transferability, and has potential to map SDD in different years with good quality of Landsat top-of-atmosphere (TOA) images embedded in Google Earth Engine (GEE). The unified model was successfully calibrated (n = 3586 data points, R2 = 0.84, MAPE = 29.8%) against SDD measured in 2235 lakes across the world, and the validation (n = 1779, R2 = 0.76, MAPE = 38.8%) also exhibited stable performance. The unified model was tuned to historical SDD measurements coincident with different Landsat sensors (L5-TM, L7-ETM+, L8-OLI) launched over the past four decades (1984-2020), thus confirming its temporal stability. Global SDD was mapped using GEE with OLI TOA products mainly acquired in 2019 to examine the spatial variation of lake water clarity (lake surface area ≥ 1 ha) all over the world. Worldwide, lake water clarity averaged 3.13 ± 1.71 m in 2019, but exhibited remarkable spatial variability due to catchment hydrological and landscape settings, lake morphology, elevation and anthropogenic impact. Inland waters in Europe (4.18 ± 1.82 m) and North America (3.84 ± 1.77 m) had the highest clarity due to greater water depth combined with less human disturbance in the high latitude regions. Lakes in South America (2.50 ± 2.33 m), Asia (2.44 ± 1.63 m) and Africa (2.36 ± 0.72 m) displayed intermediate clarity. Lakes in Oceania (1.97 ± 1.48 m) exhibited the lowest clarity for all continents except Antarctica. Further, we used the mapped SDD to evaluate water trophic status using the Carlson trophic state index. Our results indicate that, in 2019, about 63.6% of the lake areas and 47.8% of total lake numbers (2,219,627/4,646,056) were oligotrophic for global lakes, while about 23.6% areal percent and 37.1% of lake numbers are eutrophic mostly as a result of their being located in agricultural and urban-dominated drainage basins. This study, for the first time, provides water clarity information for lakes with area ≥ 1 ha all over the world with 30-m resolution and facilitates the understanding of the water clarity relevant to TSM (r = 0.95), Chl-a (r = 0.73), total phosphorus (r = 0.75), total nitrogen (r = 0.60), which could further provide water clarity data and technical support for trophic level evaluations as well. This unified model could serve as a powerful research tool for long-term monitoring of aquatic ecosystems and assessing their resilience to anthropogenic disturbance and climate change-related stressors.