CCSS: Reference-free and Spatial-aware Deep Sensor Array Decoding towards High-fidelity Remote Health Monitoring

远程健康监测通过方便和引人注目地跟踪人们的健康状况，在大数据驱动的精准医疗方面非常有希望。然而，当传感设备被放置在身体之外进行远程监测时，捕获的人体信号通常非常微弱。这是因为当信号从人体传播到设备时，它会迅速衰减。此外，如果环境中有多于一个人，则目标人的信号可能会受到干扰。针对这些关键挑战，该项目将通过创新遥感信号传感和解码系统架构，推动高保真远程健康监测的科学研究。该项目将极大地推动全民健康朝着普适、高保真、长期大数据建设的方向发展。更具体地说，本项目将设计一种新型的深度传感器阵列译码系统，该系统利用数据驱动的深度学习算法来译码噪声和弱信号，而不需要用于传播诱导失真估计的参考信号。此外，深度学习将利用多传感器的空间信息来提高信号的保真度，并从噪声和干扰中恢复感兴趣的信号。该项目将通过新课程开发、新的教学实践、课程改进和广泛的学生培训，进一步促进研究-教育一体化。PI将继续扩大不同背景的本科生、女性和少数民族学生以及K-12学生的参与，从而有效地培养下一代工程师和研究人员。该项目将创新一种新型的深度传感器阵列解码系统，可以从远程捕获的嘈杂和微弱的信号中解码感兴趣的信号，朝着有望实现远程健康监测和精准医学大数据的方向发展。利用深度学习算法对传感器阵列采集的多传感器信号进行分析，以学习噪声模式，抑制噪声，并对高保真信号进行解码。这种数据驱动的方法不需要通常用于传播引起的失真估计的参考信号，从而实现了智能和方便的信号解码。传感器阵列捕获的空间动态信息编码了感兴趣信号的复杂信息，通过深度学习支持的信号解码可以有效地学习。此外，如果环境中有多个人，深度学习算法将学习分离感兴趣信号。深度学习算法将针对目标用户的特定信号模式进行学习，并利用深度学习算法从捕获的多传感器信号中挖掘与目标相关的模式。提出的系统架构将通过真实世界的实验进行进一步的评估，以展示系统的通用性创新和有效性。新的系统架构将广泛促进各种远程健康监测应用，通过普遍和方便的健康大数据建立促进国民健康，并促进用于高保真远程健康监测的深度传感器阵列解码的科学。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Remote health monitoring is highly promising for big data-driven precision medicine, through conveniently and obtrusively tracking health conditions of people. However, when the sensing device is placed off-body for remote monitoring, the captured human signal is usually very weak. This is because the signal quickly decays when it propagates from the human body to the device. Further, the signal of the target person may be interfered if there are more than one person in the environment. Targeting these crucial challenges, this project will advance the science of high-fidelity remote health monitoring, through efforts on innovating the remote signal sensing and decoding system architecture. This project will greatly advance the national health towards pervasive, high-fidelity, and long-term big data establishment. More specifically, this project will design a novel deep senor array decoding system, which leverages the data-driven deep learning algorithm to decode the noisy and weak signal, without needing a reference signal used for propagation-induced distortion estimation. Besides, the multi-sensor spatial information will be leveraged by deep learning to boost the signal fidelity and recover the signal-of-interest from noise and interferences. The project will further contribute to research-education integration through new course development, new pedagogy practices, curriculum enhancement, and broad student training. The PI will continue broadening the participation of undergraduate, women and minority students, as well as K-12 students, with diverse background, thereby effectively training the next-generation engineers and researchers.This project will innovate a novel deep sensor array decoding system, which can decode the signal-of-interest from the noisy and weak signal remotely captured, towards promising remote health monitoring and precision medicine big data. The multi-sensor signal captured by a sensor array, will be analyzed by the deep learning algorithm to learn the noise patterns, suppress the noise, and decode the high-fidelity signal. This data-driven approach does not need the reference signal that is usually used for propagation-induced distortion estimation, thereby enabling intelligent and convenient signal decoding. The spatial dynamics captured by the sensor array encode complex information about the signal-of-interest, can be effectively learned with the deep learning-empowered signal decoding. Besides, the deep learning algorithm will learn to separate the signal-of-interest if there are more than one person in the environment. The specific signal patterns for the target user will be learned and used by the deep learning algorithm to mine the target-relevant patterns in the multi-sensor signal captured. The proposed system architecture will be further evaluated with real-world experiments, to demonstrate the generalizable innovation and the effectiveness of the system. The novel system architecture will broadly contribute to various remote health monitoring applications, advance national health with pervasive and convenient big health data establishment, and promote the science on deep sensor array decoding for high-fidelity remote health monitoring.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.