Biomedical signal quality analysis for wearable technologies

点击翻译按钮获取中文摘要

Advances in technology have led to new biomedical monitoring devices, including wearables, which allow for continuous, long-term, and ambulatory monitoring, along with the potential for real-time feedback. Wearables have broad applications, including health and wellness, fitness, human-computer interaction, and rehabilitation. The wearables market is rapidly growing (37% compounded annual growth) with device revenues reaching almost $100 billion by 2021. Unlike conventional biomedical monitoring, performed by experts in well-controlled environments, wearables are often used by non-expert end-users in a diverse set of uncontrolled environments. As a result, wearables are highly susceptible to a variety of contaminants (i.e., noise and artifacts). Contaminated biomedical data can result in misinterpretation of the data, including misdiagnoses and false alarms. Poor data quality is a significant barrier for mainstream adoption of wearables. This research program supports the development of personal biomedical devices, including wearables. In particular, the proposed research aims to develop biomedical signal quality analysis methods that will improve the accuracy and robustness of wearable devices and addresses several important research gaps. The research advances the state-of-the art, within the growing area of biomedical signal quality analysis, developing novel approaches to detect, identify, quantify, and model contaminants in data recorded by wearables. The research employs advanced machine learning approaches, including deep learning, which can offer superior performance. It also investigates the topic of biomedical signal quality analysis within a current, cutting-edge context, such as compressive sensing, which can be important for wearables at it reduces storage, bandwidth, and power requirements. The proposed research will have an important impact for wearables, addressing fundamental issues of poor data quality. For example, research outcomes can be used to enable wearable systems to disregard data segments of poor quality, avoiding misinterpretations, while retaining valid and useful biomedical data. The increased accuracy and robustness of information-leveraging methods for wearable devices is vital for the utility, user-experience, and user-acceptance of these devices.