Hardware Friendly Machine Learning Integrated Circuits and System for Low Power Wearable Wireless Electrocardiogram Sensors

未来的可穿戴无线生物医学传感器需要新的技术来克服在实现智能信号传感和处理方面日益增长的挑战，电池寿命的短缺，以及延迟和安全问题。例如，当前美国医疗保健系统的一个主要问题是，传感和处理医疗数据需要大量且昂贵的资源。为了缓解这一问题，可穿戴医疗设备有望提供对生理信号的自动监测和处理，并能够识别异常信号，并在必要时联系医疗系统。这类设备是未来“无人医疗护理系统”的关键部件。本项目的目标是通过系统的努力，最终解决下一代无线可穿戴生物医学传感器面临的挑战，包括低功耗电路设计、硬件友好算法设计和通信系统分析等方面的相关研究。该项目还包括整个可穿戴传感器的集成电路设计和表征，以及每个构建块的功率预算估计。由于高能效的智能无线设备是各种现有和新兴移动传感系统的关键组件，该项目的成果可能会对我们的生活质量产生直接的技术和社会影响。为了验证这些好处，该项目计划直接探索拟议系统在老年护理应用中的研究影响，这对新墨西哥州来说是一个特别重要的主题。除了通过拟议的研究项目培训本科生和研究生外，该项目概述的活动的教育影响还包括增加少数族裔学生的参与，并吸引高中生参加STEM大学课程。可穿戴式心电传感器是检测心律失常的重要可穿戴医疗设备之一，因为患者甚至有心脏不适的正常人都需要连续的心电监测。具有无线身体传感器网络的可穿戴式心电传感器是最好的候选者之一。近年来，机器学习已成为一种很有前途的解决方案，并已被应用于生理信号的连续监测以进行传感器处理。由于延迟、安全和隐私要求，医疗设备首选传感器处理，而不是将原始数据发送到云。因此，在可穿戴传感器应用中，能够适应实时处理而不需要太多数据存储和移动的机器学习算法是首选。该项目解决了可穿戴无线生物医学传感器，特别是心电传感器中上述技术挑战的基本计算问题。我们的目标是找到一种替代的传感和处理电路体系结构，以实现具有本地处理能力的功率受限无线传感器的低计算开销机器学习算法。这项研究旨在以跨学科的方式采用集成电路、硬件友好的机器学习算法和高能效的无线通信系统的想法，显著推动低功率无线可穿戴生物医学传感器架构的最先进水平。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。

Future wearable wireless biomedical sensors demand novel technologies to overcome the increasing challenge in implementing intelligent signal sensing and processing, the shortage of battery lifetime, as well as latency and security issues. For instance, one major problem with the current U.S. health care systems is that sensing and processing medical data require significant and costly resources. To alleviate this problem, wearable medical devices are expected to provide automatic monitoring and processing of physiological signals and be capable of identifying abnormal signals and contacting medical systems if necessary. Such devices are the key components in the future "unmanned medical nursing systems". The goal of this project is to ultimately address the challenges of next-generation wireless wearable biomedical sensors by systematical efforts, which include interrelated studies in low power circuit design, hardware-friendly algorithm design, and communication system analysis. The project also covers the integrated circuit design and characterization of the overall wearable sensor with the power budget estimation of each individual building block. Since the power-efficient smart wireless device is a critical component for a wide range of existing and emerging mobile sensing systems, the outcomes of this project can result in a direct technological and societal impact on the quality of our lives. To validate the benefits, the project plans to directly explore the research impact of the proposed systems in elderly care applications, which is a particularly important topic for the state of New Mexico. In addition to training undergraduate and graduate students via the proposed research projects, the educational impact of the activities outlined in this project includes increasing participation of minority students and attracting high school students to STEM college programs. Wearable Electrocardiogram (ECG) sensors are one of the important wearable medical devices for arrhythmia detection, as continuous ECG monitoring is needed by patients and even by normal people with uncomfortable heart feelings. A Wearable ECG sensor with wireless body sensor networks is one of the best candidates. Recently, machine learning has become a promising solution and has been applied to continuous monitoring of physiological signals for on-sensor processing. Due to latency, security, and privacy requirement, on-sensor processing rather than sending the raw data to the cloud is preferred in medical devices. Therefore, a machine learning algorithm that can accommodate real-time processing without too much data storage and movement is preferred in wearable sensor applications. This project addresses the fundamental computing issue of the above-mentioned technical challenges in wearable wireless biomedical sensors, especially ECG sensors. The goal is to find an alternative sensing and processing circuit architecture to enable low-computing overhead machine learning algorithms for power-limited wireless sensors with local processing capability. The research aims to significantly advance the state-of-the-art in low-power wireless wearable biomedical sensor architecture by employing ideas in a cross-disciplinary fashion from integrated circuits, hardware-friendly machine learning algorithms, and power-efficient wireless communication systems.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.