Multimodal Sensor Fusion and Bio-Signal Processing for Vital Sign Estimation (UNOSECO)

用于生命体征估计的多模态传感器融合和生物信号处理（UNOSECO）

• 批准号: 313380423
• 负责人: Professor Dr.-Ing. Steffen Leonhardt
• 金额: --
• 依托单位: Lehrstuhl für Medizinische Informationstechnik (MedIT)
• 依托单位国家: 德国
• 项目类别: Research Grants
• 财政年份: 2016
• 资助国家: 德国
• 起止时间: 2015-12-31 至 2020-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/313380423?language=en
• 关键词: Multimodal; Sensor; Fusion; Bio; Signal

The goal of this project is the investigation of methods of multimodal bio-signal processing for vital sign estimation. The focus lies on the development of a methodology that is independent of the sensing method and especially considers non-contact and unobtrusive modalities, which in general are very susceptible to disturbances. However, in contrast to traditional clinical instrumentation they can potentially be integrated into objects of everyday life. For this, a unified, macroscopic model of bio-signal generation is necessary. Our approach proposes the generation of cardiac signals from a virtual source. This virtual source signal is transmitted through virtual channels, at the end of which a real physical signal such as electrocardiogram, photoplethysmogram or ballistocardiogram can be measured. The first goal of this project is answering the question which assumptions have to be made about such a source-filter model for the generation of cardiac and later even respiratory signals. Based on these assumptions, existing signal processing methods will be evaluated and adapted or created from scratch. If the proposed source-filter model is adapted for each subject, two advantages arise: First, model-based sensor-fusion becomes possible. Second, diagnostic information can potentially be extracted from the adapted model parameters. To validate the proposed model and to test signal processing methods, a simulation environment for multimodal bio-signal and their distortions needs to be created. Additionally, real measurement data needs to be recorded. For this, several unobtrusive sensing modalities already available at the institute need to be integrated into a unified multimodal measurement setup. In addition to the recording of clean data, this setup will allow the generation of various, reproducible artifacts. The recorded measurements and reference data will be provided to the public via UNOVIS, the "MedIT Public Unobtrusive Vital Sign Database". Hence, the overall goal of this project is the generation of general methods and structures to advance research in the area of unobtrusive vital signs estimation. In subsequent projects, these methods and structures can be applied to different areas of applications and pathologies.

本课题的目标是研究多模态生物信号处理的方法，以估计生命体征。重点在于开发一种独立于传感方法的方法，特别是考虑非接触和不显眼的模式，这些模式通常很容易受到干扰。然而，与传统的临床仪器相比，它们可以潜在地集成到日常生活的物体中。为此，有必要建立一个统一的、宏观的生物信号产生模型。我们的方法提出了从虚拟源产生心脏信号。这个虚拟源信号通过虚拟通道传输，在通道的末端可以测量真实的物理信号，如心电图、光电容积描记图或心电图。这个项目的第一个目标是回答这样一个问题：对于产生心脏甚至呼吸信号的源滤波器模型，必须做出哪些假设。基于这些假设，现有的信号处理方法将被评估和调整或从零开始创建。如果提出的源滤波器模型适用于每个主题，则有两个优点：首先，基于模型的传感器融合成为可能。其次，可以从适应的模型参数中提取诊断信息。为了验证所提出的模型和测试信号处理方法，需要创建一个多模态生物信号及其畸变的仿真环境。另外，需要记录真实的测量数据。为此，该研究所已有的几种不显眼的传感模式需要集成到统一的多模态测量装置中。除了干净数据的记录之外，这种设置还允许生成各种可重复的工件。记录的测量数据和参考数据将通过联科观察处，即“医疗信息公共不显眼生命体征数据库”向公众提供。因此，该项目的总体目标是生成一般方法和结构，以推进非突发性生命体征估计领域的研究。在后续的项目中，这些方法和结构可以应用于不同的应用领域和病理。