A Review on Multisensor Data Fusion for Wearable Health Monitoring
Arlene John, Barry Cardiff and Deepu John · 2024
Read A Review on Multisensor Data Fusion for Wearable Health Monitoring by Arlene John, Barry Cardiff and Deepu John (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-05895.txt.
Introduction
Arlene John, Barry Cardiff, and Deepu John review how data from many sensors can be merged for health tracking. Their review appeared in two thousand twenty-four. Wearable devices now watch the body around the clock without breaking the skin. Each sensor alone is noisy or partial. Fusion methods combine their streams into a fuller picture. The review maps how this field has grown.
The authors set out shared terms and classes used in fusion work. They cover the main algorithms, including Kalman filters and related tools. Uses in defense, self-driving cars, robotics, and image work give context. The heart of the review is health monitoring systems. It weighs key methods, open challenges, and likely next steps. Builders of wearable health tools are the main audience.
Catalog record
| CREATOR | Arlene John, Barry Cardiff and Deepu John |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 95 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-05895
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Review on Multisensor Data Fusion for Wearable Health Monitoring Arlene John, Barry Cardiff and Deepu John Abstract The growing demand for accurate, continuous, and non-invasive health monitoring has propelled multi-sensor data fusion to the forefront of healthcare technology. This review aims to provide an overview of the development of fusion frameworks in the literature and common terminology used in fusion literature. The review introduces the fusion classification standards and methods that are most relevant from an algorithm development perspective. Applications of the reviewed fusion frameworks in fields such as defense, autonomous driving, robotics, and image fusion are also discussed to provide contextual information on the various fusion methodologies that have been developed in this field. This review provides a comprehensive analysis of multi-sensor data fusion methods applied to health monitoring systems, focusing on key algorithms, applications, challenges, and future directions. We examine commonly used fusion techniques—including Kalman filters, Bayesian networks, and machine learning models. By integrating data from various sources, these fusion approaches enhance the reliability, accuracy, and resilience of health monitoring systems. However, challenges such as data quality and differences in acquisition systems exist, calling for intelligent fusion algorithms in recent years. The review finally converges on applications of fusion algorithms in biomedical inference tasks like heartbeat detection, respiration rate estimation, sleep apnea detection, arrhythmia detection, and atrial fibrillation detection. I Introduction Healthcare expenditure has been drastically increasing worldwide over the past several decades due to an ageing population and chronic health conditions. Continuous and proactive monitoring of vital health signs outside a lab environment using wearable sensors is widely seen as the most attractive option for healthcare management. Non-invasive ambulatory physiological signal monitoring research focuses on monitoring primary physiological signals like heart rate, respiration rate, peripheral capillary oxygen saturation, blood pressure, etc and derived physiological parameters like drowsiness levels, gait, etc. However, there are several challenges involved in making wearable sensors a reality. One of the major challenges is the low quality of the signals acquired due to motion artifacts, lack of robustness to a node failure, power demands and architecture area. Corrupted signals can cause significant fluctuations in measured or derived parameters, resulting in false alarms in ambulatory health monitoring devices. This can lead to alarm fatigue and delay responses to critical events [ 1 ] . In other cases, a single lead may fail, resulting in a loss of information. Additionally, due to the compact design of sensors, electrode contacts are often placed much closer together than standard measurement locations. This proximity leads to low signal amplitude, further contributing to signal quality issues and causing poor feature extraction and data interpretation [ 2 ] . Data fusion has emerged as a solution that provides improved accuracy and more precise inferences compared to those derived from a single sensor source. Multi-sensor data fusion is beneficial in scenarios when one of the sensor sources utilized in a monitoring or inference task is faulty or corrupted by noise, such that the other sensors can contribute more to the inference process. Data fusion algorithms enhance the performance of a particular task by combining information from multiple sensor sources [ 3 ] . Although data fusion is a widely used solution, no definitive methods or methodologies have been established for fusing physiological signals to enhance the quality of health monitoring or inference. This review aims to condense the diverse field of data fusion to identify relevant fusion methodologies applicable to monitoring and inference, specifically for 1-dimensional time-series data obtained from health monitoring sensors, such as electrocardiograms (ECG), photoplethsymogram (PPG), accelerometer data etc. However, with this review, the expectation is that the fusion methodologies identified could also be extended to broader applications in healthcare. II Methodology The theory of data fusion is also highly relevant to fields such as autonomous vehicles, environmental monitoring, structural health monitoring of civil infrastructure, robotics, etc. In this review, we delve into the details of existing fusion frameworks in disparate fields to conclude a suitable framework classification strategy for the multisensor fusion works carried out in the field of biomedical signal processing. This review was carried out using a systematic search across multiple databases, including Scopus, Web of Science, PubMed, IEEEXplore, and Google Scholar. The keywords used included “data fusion”, “multisensor (multi-sensor) data fusion”, “fusion methodology”, and “fusion frameworks.” A snowballing search on the included articles were carried out to ensure no works were missed. The included studies are discussed in this review and are used to develop or compile a preference list of fusion frameworks that are useful in classifying multisensor fusion algorithms developed for health monitoring for biomedical 1-dimensional time series signals. After the inclusion of the articles on fusion frameworks, a similar search was carried out for the monitoring applications included in this review, such as multi-sensor data fusion for heartbeat detection, heart rate estimation, respiration rate estimation, sleep apnea detection, arrhythmia detection, and atrial fibrillation detection. The inclusion criteria is that the works should include the fusion of multiple sensors for the development of a multi-sensor fusion algorithm for these applications. Fusion of different features obtained from a single sensor source is excluded. III Data fusion frameworks in literature This section provides a brief description of data fusion architectures discussed in the literature. Fusion methods are often categorized based on the application domain, fusion objective, sensor type, sensor suite configuration, etc. A number of data fusion frameworks have been developed over the years to aid in the identification of the most appropriate fusion methodologies for varying applications. Table I depicts the development timeline of various fusion architectures and data fusion reviews. In this section, we elaborate on these fusion methodologies and present findings chronologically, helping the reader trace potential relationships and influences within the literature. The Joint Directors of Laboratories (JDL) data fusion framework, the oldest and most widely used framework, was developed to support advancements in military applications [ 4 ] . JDL introduces processes, functions, and different techniques with four key processes: object refinement, situation refinement, threat refinement, and process refinement (Fig. 1 ). While this model has been applied in large-scale military projects, it appears to be highly specific to this domain. TABLE I : Timeline of the development of fusion frameworks 1987 \cdots\cdots \bullet JDL [ 4 ] . 1988 \cdots\cdots \bullet Pau [ 5 ] . Schoess and Castore [ 6 ] . Durrant-Whyte [ 7 ] . 1989 \cdots\cdots \bullet Luo and Kay [ 8 ] . Thomopoulos [ 9 ] . 1993 \cdots\cdots \bullet Clement et al. [ 10 ] . 1997 \cdots\cdots \bullet Dasarathy [ 11 ] . 1998 \cdots\cdots \bullet Harris et al. [ 12 ] . Pohl and Genderen [ 13 ] . 2000 \cdots\cdots \bullet Bedworth and O’Brien [ 14 ] . 2001 \cdots\cdots \bullet Keithley [ 15 ] . 2003 \cdots\cdots \bullet Laskey and Mahoney [ 16 ] . Chang et al. [ 17 ] . Carvalho et al. [ 18 ] . 2004 \cdots\cdots \bullet Brooks et al. [ 19 ] . Ahmed and Pottie [ 20 ] . Blasch et al. [ 21 ] . 2005 \cdots\cdots \bullet Esteban et al. [ 22 ] . 2008 \cdots\cdots \bullet Cohen and Edan [ 23 ] . 2009 \cdots\cdots \bullet