Biomedical Signal Sensing and Analysis

点击翻译按钮获取中文摘要

Wired and wireless sensors for biomedical signal acquisition have become ubiquitous for long-term health monitoring and tele-health applications. Biomedical signal analysis which combines the aspects of signal processing and machine learning is becoming an area of importance, both for hospital-based instrumentation and for home-based medical devices. While biomedical instrumentation has improved over the years and is commonly used in health care and research, two main challenges still persist: 1) robust analysis and interpretation of the signals that are acquired, and 2) power requirements of the devices. The current biomedical devices consume large amount of power, and eventually become inefficient and non-compliance especially in long term health monitoring using wearable devices. One of the main reasons for this has been the inefficient use and design of biomedical signal processing algorithms right from signal acquisition to analysis, and decision making. In the proposed research, compressive sensing (CS) technique in which redundancy and sparsity characteristics associated with a biomedical signal are systematically exploited, will be used in designing low-power and robust data acquisition system. A new direction of using CS for analysis and interpretation of the signals will be pursued. The proposed joint compressive sensing and analysis (CSA) framework will look into information theory and optimization methods that are specifically adapted for biomedical signals to handle the variability (due to non-stationarity and non-linearity) of the signals, the inter-relationships and interactions exhibited among physiological systems, and the domain-specific physiological artifacts and interferences. The performance measures of these newly developed CSA techniques will help in improving the power and bandwidth requirements of biomedical sensors used in long term monitoring applications, and will provide robust and efficient extraction of salient features from the signals for improved decision making in healthcare. The applications of the proposed research will tremendously benefit the areas of cardiac signal analysis, neuromuscular signal analysis, respiratory signal analysis, and design of low-power wearables for health and wellness. The techniques and algorithms developed here could also be extended and adapted to other emerging areas such as biometrics/cybersecurity, RADAR/SONAR, and multimedia related signal processing and machine learning.