Research Of Wearable Vital Sign Sensing Technology For Healthcare

Public health becomes a research widely discussed topic with challenges concerning the outburst of global population. Especially the increasing population of elderly and patients with chronic disease, shortage and imbalanced distribution of medical resources, etc. As an efficient and economic technical solution, mobile digital medical, particularly the body sensor network enabled smart wearable systems, are proposed for the public health in the new century. The medical way will go through fundamental changed, and tele-health monitoring (in home health monitoring), tele-medical, tele-care can be achieved by using such technology. However, there are still some existing challenges when smart wearable systems have been used in healthcare applications, which include vital signs sensing, node energy efficiency and information fusion. To address this existing issues in smart wearable systems, the researches of wearable vital sign acquisition, energy efficient data compression for resource constrained nodes, the technology of signal processing and information fusion are proposed from the sides of sensing, transmission, and information abstraction in this study. The main works of this thesis include:(1) Research and development of a smart fabric based wearable vital signs system, and a compressed sensing (CS) based ultra-low-power unipolar ECG node. (2) Digital integrate-and-fire sampler based dynamic compression algorithm is proposed for energy efficient wireless ECG bio-sensor. (3) Information enhanced sparse binary matrix (IESBM) is proposed for ECG compressed sensing measurement matrix. (4) Priori-block sparse Bayesian learning (P-BSBL) algorithm is proposed for CS based ECG recovery. (5) Modified frequency slice wavelet transform (MFSWT) is proposed for bio-signal processing. (6) Unsupervised feature extraction and deep architectural model is proposed for automatic heartbeat classification. The progresses of our research include wearable ECG measurement by using dry fabric electrodes where the sensor nodes of system are well organized by using the proposed ring structure fabric communication bus. The advantages of fabric communication bus include plug and play; low power consumption; the ECG node recognizes ECG measurement form unipolar human body surface. The advantages of ECG node include low price, small size, light weight and low power. The dynamic compression algorithm and CS based vital signs compression algorithms reduce the amount of transmitted data, enhance energy efficiency and prolong the lifetime of the node. Their common advantages include low computational complexity, low resource consumption and small recovery distortion. Particularly, the lifetime of dynamic compression algorithm implanted MICAz node prolong 76.3%, with PRD<9% and the lifetime of CS based Micaz pulse node prolonged 54.20% with average PRD is 4.23%. The IESBM reduces the overall recovery ECG distortion and the areas of interested and the P-BSBL algorithm efficient reconstructs no sparse ECG signal. Finally, the signal components are precisely located in the time-frequency spectrogram by using MFSWT. Thus, the proposed method based on supervised feature extraction and deep architectural model achieves high accuracy normal beats (n), ventricular ectopic beats (SVEB, S), ventricular ectopic beats (VEB, V) and fusion beats (f) classification, where the average sensitivity is 97.76%, g-mean is 88.69%. These results indicate that the proposed method have further enriched the wearable vital signs sensing technology, and provide valuable theoretical and technical support for vital signs sensing, long-term chronic disease monitoring, disease prevention, and intelligent assessment diagnosis.