Data acquisition optimization for smart monitoring networks

点击翻译按钮获取中文摘要

Data Acquisition Systems (DAS) digitize analog signals in order to be digitally processed by embedded systems in monitoring or control applications. DAS are at the front-end of the chain of data processing and, as such, any quantization and sampling error that is input to the computing system may cause catastrophic failures in the correct interpretation of the acquired signals. Current DAS have no provisions for Measurement Quality Assurance (MQA) to certify in real-time the correctness and trustworthiness of measurements by providing metadata on the precision, accuracy, conditions and utility of measurements, and which are essential requirements for operation in remote unsupervised environments. The lack of this information made impossible the development of viable monitoring technologies for personal healthcare applications or constrained their proliferation in other domains. Losing data because of unreliable operation, due to scarce resources caused by unexpected events, fuelled a lack of confidence in these technologies, as well. In order to address the resources efficiency, reliability and versatility, while increasing accuracy, trustworthiness and adequacy of the collected information, we propose to conceiving, developing and optimizing architectures, algorithms and models at different levels of data acquisition. The quantization process will be marked by MQA qualifiers which will objectively and globally characterize the measurands (acquired signals), measurement uncertainty, acquisition conditions (compliance with the requirements of standards), and measuring devices (Hardware and Software), including the accuracy of the indirect measurement algorithms. This research will employ compressed sensing techniques in order to reduce digitization relative errors and optimize the density of transmitted data. Floating-point differential quantization methods will be used to insure a broad dynamic range of the acquired signals and eliminate redundant information. Adaptive sampling rate algorithms will be studied in the context of the acquisition of sparse signals with strong deterministic components, but which may occasionally deviate from stable operation. Adaptive compressive sampling and quantization will mitigate the power consumption if resources get scarcely, while maximizing the data acquisition capacity and minimizing the measurement uncertainty. One of the most challenging applications of DAS is the non-invasive personal health monitoring. Assessment of quotidian physiological parameters relies on measurement and processing of bio-signals that are strongly correlated with these parameters and which can be acquired non-invasively. Optimization of methods in estimation of physiological parameters based on multi-sensors and algorithms fusion will be a central objective of this research plan. The amount of collected data will be optimally adapted to the resources constraints. Current technologies require more power for data transmission rather than processing. Hence, the local DAS management will decide to calculate locally the parameters whenever is possible, to minimize communication, in order to save energy. This research project will be validated by implementing its findings on a multi-sensor DAS platform that will be capable of fusing data, adapting its operation and structure to the dynamics of the monitored signals, selftesting, self-calibrating and estimating the MQA parameters. This project aims at improving the quality of health services in Canada by enabling people to monitor and report their health conditions. Not only will medical practitioners have access to trustworthy data, but governmental agencies will also be able to concurrently monitor the efficiency of their policies and investments.