Human states are fundamentally influenced by inherent physiological signals and are manifested through physical body statuses. Accurate, continuous monitoring of human physical and physiological signals is critical for enhancing healthcare, personalizing education, and facilitating human interaction with the physical environment. However, current methods for acquiring human data often require heavy instrumentation of the environment, demand extensive manual inputs, or rely on wearable sensors that are often rigid or adhesive, leading to discomfort over prolonged use.
This thesis presents the transformation of everyday conductive fabrics into a natural, pervasive sensing platform capable of detecting human physical body movements and underlying physiological signals. Using fabrics as sensors for reliable human sensing presents numerous practical challenges, including motion noise caused by the fabric’s intrinsic flexibility, data variability due to individual and environmental differences, and the limited computing power of wearable computing units. This thesis addresses these challenges by combining novel hardware and systems techniques with efficient computational methods.
The first part of the thesis demonstrates fabric-based physical sensing by developing a sleeve embedded with sensing fabrics to reconstruct elbow joint motion. Our design combines conductive, stretchable fabrics that sense strain with pressure fabrics that sense pressure during elbow motion. We develop biomechanics-inspired algorithms for joint angle reconstruction and construct mathematical models employing normalization strategies to achieve generalization across different conditions. These design elements address the fabric’s non-monotonic response to motion, its sensing instability, and user diversity. Such systems can facilitate patient rehabilitation and motion teaching.
The second part of the thesis focuses on fabric-based physiological sensing, where we repurpose conductive fabrics as sensing electrodes for measuring biopotential signals on the human body, such as electrocardiogram (ECG) and electromyography (EMG). We optimize sensor designs to overcome the weak signal-to-noise ratio of these biopotentials and mitigate motion artifacts. We also apply neural networks for denoising and identifying user-independent features. We developed a fabric necklace made of these fabric sensors to monitor the practice of Kangaroo Mother Care. The system senses ECG transmission across the human body to infer the duration of chest-to-chest skin contact between the infant and caregiver, as well as the infant’s vital signs, all of which hold significant clinical value. Our latest effort embeds fabric sensors into pillowcases to sense ECG and EMG around the user’s neck and infer multiple sleep variables (e.g., sleep pose, vital signs). This enables non-intrusive, large-scale sleep monitoring without disrupting normal sleep routines.
Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/7pz8-gn75 |
Date | January 2024 |
Creators | Shao, Qijia |
Source Sets | Columbia University |
Language | English |
Detected Language | English |
Type | Theses |
Page generated in 0.1771 seconds