THE MAIN METHODOLOGICAL FEATURES OF FACE RECOGNITION
The paper analyzes the main methodological features associated with recognition. The similar approaches are used in systems of security, protection. There have been three basic steps, combining different approaches to face detection. The key components of the procedures of recognition are: implementation of face detection, the keypoints individuals, view individuals as vectors of features. Provided the complexity of the allocation of persons in video data. The analysis of perspectives of application of algorithms of Viola-Jones and Dalal-Triggs. It is noted that the model of the deformable parts allows the use of a strong low-level characteristics based on histograms of oriented gradients, like the algorithm Dalal-Triggs.
Keywords: face recognition, algorithm, digital image processing.
THE ALGORITHMS FOR FACIAL RECOGNITION
The basic idea of face recognition is the selection of informative features in the face image, encoding, and comparison of the encoded entity with the database. In this paper the analysis of algorithms based on the method of principal components, linear discriminant analysis, detection of local features, with the application of Gabor wavelets, discrete cosine transform, local binary patterns are given. It is noted that the correlation methods are characterized by computational complexity and require large amounts of memory, in this regard, in practice it is reasonable to use appropriate methods to reduce the dimensionality of the features. Shows the latest developments of the company “Vokord” based on the use of deep neural networks using a test database with a million photos.
Keywords: detection, face, algorithm, machine vision, safety.
THE FEATURES OF MODERN METHODS OF ESTIMATION AND FILTERING OF SIGNALS
The paper considers the peculiarities of signal processing in radar systems based on the methods of filtering and parameter estimation. The conditions that are required in order to estimate the parameters of the signals, is usually performed for stage processing of radar signals, which is based on special signal processors are shown. The characteristics of primary and secondary radar data are marked. In the models presented on the basis of stochastic differential equations that provide: models associated with the continuous change of the parameters and models associated with the continuous quasi-linear filtering. The analysis of models related to the estimation of the parameters of intense maneuvering objects, which did not provide a clear transition to the algorithms related to nonlinear filtering.
Keywords: signal, filtering, estimation, radar, electronic means.
THE PROPERTIES OF PARALLEL COMPUTING SYSTEMS
This paper discusses the basic properties of parallel systems. The approaches, which can be used to achieve parallelism are specified. The characteristics of Flynn’s taxonomy, which describes the interaction of flows of commands and data streams are shown. A classification of multiprocessor computing systems is given. The architecture of multiprocessor systems with shared memory is shown. The features of systems with heterogeneous shared memory are specified. The description of the architecture of computer systems having distributed memory is carried out. The differences of multicomputers from multiprocessors are discussed. The advantages of clusters as a means of parallelization are shown.
Keywords: parallel computing, CPU, network, and data flows.
THE METHODS OF PROCESSING OF IR-IMAGES WITH DISTORTION AND NOISE
In this paper the analysis of algorithms related to automatic recognition on the basis of opto-electronic systems is carried out. As a generalized criterion of efficiency in the monitoring of information systems use parameters such as time and reliability of forecasts for the development of observed situations or processes is shown. It is noted that it is necessary to release operators from routine operations and to bring to them the necessary and sufficient variety of data and forecasts on the basis of which they the importance of reliability to make decisions for a wide range of changing conditions of observation. Based on quantitative data there are opportunities for short-term forecasting of the maneuvers of the objects of accounting of the environment for different wavelength ranges, to compensate for noise using two sources of signal for the correction: transducer, having a known temperature and a signal of the scene.
Keywords: IR-image, distortion, noise, signal processing.
THE FEATURES OF AUTOMATIC SPEECH RECOGNITION
In the paper the basic of speech acoustic and actual items in the field of speech recognition. It is noted that the difficulty of recognition of speech signals associated with their probabilistic properties and diversity impact of interference. Users need to pre-configure systems on their voice. The system factors are shown on the example of acoustics features of speeches and the system of fonetic signs.
Keywords: the speech recognition system of recognition, handling of signals, disturbance.
THE SIMULATION IN THE TIME DOMAIN FOR SCATTERING OF RADIO WAVES ON THE OBJECT LOCATED UNDER THE GROUND COVER
V.N. Kostrova, E.A.Rybalchenko
The paper discusses an analytical solution in the time domain Pocklington equation for a straight thin wire of finite length under land cover at a certain depth. This wire is excited by an electromagnetic wave. The description of the transition surface to air is carried out on a simplified model for the reflection coefficient. The analytical solution is based on the use of Laplace transform and Cauchy’s theorem. The results of numerical calculations of the excited current in the time domain for different parameters – the conductivity of the medium are given. Recommendations on the use of the method of parallel computing to accelerate numerical solution of the problem are shown.
Keywords: communication, scattering of electromagnetic waves simulation in time domain integral equations.
MODELING OF RADIO PROPAGATION IN INDOOR-BASED FINITE-DIFFERENCE METHOD IN TIME DOMAIN
D.G. Panarin, A.V.Danilova
The paper discusses the features of the solutions of Maxwell’s equations in the time domain. The authors discussed the use of finite difference method in time domain to evaluate the extent of signal attenuation during its propagation inside the building. Due to the fact that there are restrictions on the amount of available memory, we carried out simulations for two-dimensional problems, which dealt with the vertical polarization of the electromagnetic wave. The results of numerical calculations demonstrating the efficiency of this method. The results can be useful in determining the optimal location of the antenna to provide required signal level at any distances, and also to explore the patterns of signal attenuation in a statistical analysis.
Keywords: wave propagation, telecommunication, method of moments, the disturbance attenuation.