Estimation of directions-of-arrival (DOA) is an important problem in various applications and a priori knowledge on the source location is sometimes available. To exploit this information, standard methods are based on the orthogonal projection of the steering manifold onto the noise subspace associated with the a priori known DOA. In this paper, we derive and analyze the Cramer-Rao bound associated with this model and in particular we point out the limitations of this approach when the known and unknown DOA are closely spaced and the associated sources are uncorrelated (block-diagonal source covariance). To fill this need, we propose to integrate a priori known locations of several sources into the MUSIC algorithm based on oblique projection of the steering manifold. Finally, we show that the proposed approach is able to almost completely cancel the influence of the known DOA on the unknown ones for block-diagonal source covariance and for sufficient signal-to-noise ratio (SNR).
The concept of Digital Twin (DT) seems promising to improve the management of patient pathways in Emergency Departments (EDs). This article proposes an agent-based architecture of a DT designed for that purpose. The core of this DT is its Information System (IS), which is regularly synchronised on the IS of the Physical Twin (PT). The agents model the ED’s resources (equipment and staff) and patients in the DT and update this information in the DT’s IS. This article shows how such a DT may operate in three modes: (0) “Digital Shadow” to monitor the ED’s current state in real time, (1) “Synchronised DT” to monitor the ED’s current and future states according to a predictive simulation, and (2) “Exploratory DT” in order to perform Monte Carlo simulations of various future states. Mode (1) is the main contribution. This proposition is illustrated in a simulation of the ED in order to demonstrate the capabilities.
22 August 2014 in IEEE Journal of Biomedical and Health Informatics
When epidemics occur, such as is the case for bronchiolitis in the Pediatric Emergency Department (ED), the patient flow in the ED incontestably increases and can lead to crowding. We bypassed this difficulty of forecasting patient flow with aggregated weekly or monthly data by tackling the problem from a different point of view. We used daily data to build a multiperiod Serfling-based model. This model is hereinafter assimilated to normal ED flow. We then used the fourth-order moment of distribution of the time series, obtained from the difference between the estimated model and the real data, to provide an early index announcing abnormal ED patient flow. This index is parameter-dependent and we provide criterion to assist in selecting the optimal parameters. A simple program based on this methodology has been developed and has been given to the pediatric physicians for testing. Thanks to this index, the Pediatric ED was able to anticipate crowding almost three weeks before the height of the bronchiolitis epidemic.
21 November 2018 in IEEE Journal of Biomedical and Health Informatics
Excessive admissions at the Emergency Department (ED) is a phenomenon very closely linked to the propagation of viruses. It is a cause of overcrowding for EDs and a public health problem. The aim of this work is to give EDs' leaders more time for decision making during this period. Based on the admissions time series associated with specific clinical diagnoses, we will first perform a Detrended Fluctuation Analysis (DFA) to obtain the corresponding variability time series. Next, we will embed this time series on a manifold to obtain a point cloud representation and use Topological Data Analysis (TDA) through persistent homology technic to propose two early real-time indicators. One is the early indicator of abnormal arrivals at the ED whereas the second gives the information on the time index of the maximum number of arrivals. The performance of the detectors is parameter dependent and it can evolve each year. That is why we also propose to solve a bi-objective optimization problem to track the variations of this parameter.
International audienceDiagnosing defects on rotating machines can be reached by several angles. When dealing with asynchronous motor drive, such physical elements rotate that a natural angle for treating the healthiness of the motor can be obtained by the use of spectral analysis tools. It is now stated that electrical or mechanical defects, which appear periodically as well, can be retrieved by analyzing the amplitude of particular frequencies inside an estimated power spectrum. When dealing with broken rotor bars detection it is essential to accurately localize the frequencies related to the slip inside the power spectrum. The diagnosis is thereafter made by indicators given with respect to their power. For actual low level of load operations, the supply frequency generally masks the frequencies which could be exploited for indicators. Therefore, we propose to cancel, as well as possible, the contribution of this supply frequency to develop the useful and closely components. The resolution should be very thin for the components to be estimated. In consequence, we use a prior-knowledge subspace-based frequency estimator, already developed in the literature, we complete with an Oblique Projection coupled with a Total Least Squares solution for estimating the power of the resulting estimated frequencies. Finally, we show by means of a real application how it contributes to improve the power spectrum estimation when compared to the FFT or periodogram-based analysis and how the aforementioned power spectrum makes the diagnosis indicator of rotor bars efficient