Yuriy S. Shmaliy

System fusion in passive sensing using a modified hopfield network

Abstract We address a new approach to the problem of improving the quality of remote-sensing images obtained with several passive systems, in which case we propose to exploit the idea of neural-network-based imaging system fusion. The fusion problem is stated and treated as an aggregate inverse problem of restoration of the original image from the degraded data provided by several image-formation systems. The non-parametric maximum entropy regularization methodology is applied to solve the restoration problem with the control of balance between the gained spatial resolution and noise suppression in the resulting image. The restoration and fusion are performed by minimizing the energy function of the multistate Hopfield-type neural network, which integrates the model parameters of all sensor systems incorporating a priori and measurement information. Simulation examples are presented to illustrate the good overall performance of the fused restoration achieved with the proposed neural network algorithm.

Matching Confidence Masks with Experts Annotations for Estimates of Chromosomal Copy Number Alterations

Structural aberrations (SAs), gains or losses in large segments of genomes, are associated with several genetic disorders. The SAs are commonly called the copy number alterations (CNAs) and their identification/classification is required to identify diseases. Many methods have been proposed to estimate the breakpoints and segmental constants in the CNAs with highest precision using the most powerful technologies of hybridization. However, locations and lengths of CNAs estimated using well-elaborated methods are often contradictory due to extensive variability of measurements and performance of the algorithms. Still much less attention is given to the estimation accuracy and it is difficult to select the best estimator. In this work, we propose to modify the confidence masks replacing the skew Laplace distribution with the asymmetric exponential power distribution (AEP) to approximate the jitter distribution in CNAs. Next, the estimates obtained using different algorithms are matched with the annotations made by experts employing the improved masks. Finally, we specify the match confidence probability of each CNAs detector algorithm respect the experts estimates.

UFIR Filtering for GPS-Based Tracking over WSNs with Delayed and Missing Data

In smart cities, vehicles tracking is organized to increase safety by localizing cars using the Global Positioning System (GPS). The GPS-based system provides accurate tracking but is also required to be reliable and robust. As a main estimator, we propose using the unbiased finite impulse response (UFIR) filter, which meets these needs as being more robust than the Kalman filter (KF). The UFIR filter is developed for vehicle tracking in discrete-time state-space over wireless sensor networks (WSNs) with time-stamped data discretely delayed on -step-lags and missing data. The state-space model is represented in a way such that the UFIR filter, KF, and filter can be used universally. Applications are given for measurement data, which are cooperatively transferred from a vehicle to a central station through several nodes with -step-lags. Better tracking performance of the UFIR filter is shown experimentally.

Fast Kalman-like optimal FIR filter for time-variant systems with improved robustness

In this paper, a fast Kalman-like iterative OFIR algorithm is proposed for discrete-time filtering of linear time-varying dynamic systems. The batch OFIR filter is re-derived in an alternative way to show that this filter is unique for such systems. A computationally efficient fast iterative form is found for the OFIR filter using recursions. It is shown that each recursion has the Kalman filter (KF) predictor/corrector format with initial conditions specified via measurements on a horizon of N nearest past points. In this regard, the KF is considered as a special case of the iterative OFIR filtering algorithm when N goes to infinity. Applications are given for the 3-state target tracking and three-degree-of-freedom (DOF) hover system. It has been shown experimentally that the proposed iterative OFIR algorithm operates much faster than the batch OFIR filter and has the computational complexity acceptable for real-time applications. It has also been demonstrated by simulations that an increase in the number of the states results in better robustness of the OFIR filter against temporary model uncertainties and in higher immunity against errors in the noise statistics.

Photonic junction based on dark spatial solitons

We present a fundamental study on the capability of a crossing of two optical waveguides based on dark-spatial solitons to act as a controllable optical beam splitter. Our study is based on the fact that the guided beam is diffracted at the waveguide crossing by an effective phase screen formed by the soliton collision profile. We find that when each dark soliton is immersed into its own bright background, the corresponding optical junction can not operate. This is because the finite width of the backgrounds acts as a low-pass filter over the diffracted beam, and because the onset of the cross-phase modulation instability effect occurs for small enough collision angles.