Heart diseases is the world’s principal cause of death, and arrhythmia poses a serious risk to the health of the patient. Electrocardiogram (ECG) signals can be used to detect arrhythmia early and accurately, which is essential for immediate treatment and intervention. Deep learning approaches have played an important role in automatically identifying complicated patterns from ECG data, which can be further used to identify arrhythmia. In this paper, deep-learning-based methods for arrhythmia identification using ECG signals are thoroughly studied and their performances evaluated on the basis of accuracy, specificity, precision, and F1 score. We propose the development of a small CNN, and its performance is compared against pretrained models like GoogLeNet. The comparative study demonstrates the promising potential of deep-learning-based arrhythmia identification using ECG signals.
22 June 2020 in IEEE Transactions on Industrial Informatics
The prodigious amount of data movement among sources, data centers, or processing elements precludes the utilization of least-squares (LS) and fusion-center (FC) based modeling and control. The LS methods are offline in nature, hence may face difficulty in real-time implementation. FC-based methods are non-robust due to single point failure, require large communication bandwidth and computationally fast processing unit. To curb these limitations, this article identifies distributed parameter systems by estimating the parameters of spatio-temporal Volterra model using in-network data processing. It can handle the immense volume of data by distributing the processing tasks of FC among the wireless sensor network nodes. To facilitate distributed optimization, the global objective function is reformulated as a multiple constrained separable problem which is then decomposed into augmented Lagrangian form. Then, alternating direction method of multipliers along with coordinate descent method is employed to obtain the global optimal solution collaboratively. Further, a communication-efficient algorithm is designed for the proposed approach to deploy in an ad-hoc network. Simulations are carried out on two industrial distributed parameter systems (catalytic rod and tubular reactor) to illustrate the practicality of the proposed algorithm.
Nonlinear problems have drawn the attention of many researchers, engineers and scientists as most of the real systems are inherently nonlinear in nature. These systems widely exist in many industrial processes such as thermal process, fluid process, pH neutralization process, and so on. They can be customarily classified into lumped parameter systems (LPSs) and distributed parameter systems (DPSs). Modeling of these systems is quite arduous, but is essential for prediction, control, and analysis. To model and analyze these nonlinear systems, many theoretical and practical methods including Volterra and block-structured models have been developed. The main objective of this paper is to concisely present the Volterra and block-structured modeling approaches along with their vital applications to nonlinear system identification and control. A much-needed effort is given in describing various modeling methods for unknown DPSs. As DPSs are of infinite dimensionality, hence they require model reduction (MR) techniques for practical implementation. Various types of MR techniques are studied in this paper where the efficient selection of spatial basis-functions is often required. Further, the adaptive modeling methods for LPSs and DPSs are studied to inculcate the benefits of adaptiveness. Finally, wireless sensor network-based distributed consensus modeling of nonlinear DPSs is discussed, diminishing the limitations of least-squares and centralized adaptive methods.
This study provides a recursive parametric identification scheme for a liquid-saturated steam heat exchanger system. The recursive identification scheme uses block-structured Wiener and Hammerstein models as model structure and recursive least squares estimation scheme as the parameter estimation method. The estimated block-oriented model provides higher accuracy of estimation than linear models provided in the literature. From the simulation results, it is observed that the Wiener model can provide 88% goodness-of-FIT, whereas Hammerstein model can provide 96% goodness-of-FIT using the said technique.
24 May 2019 in Journal of Circuits, Systems and Computers