Search

Search Results

• ALGORITHM FOR TRAINING DATA PREPARATION OF CONVOLUTIONAL NEURAL NETWORKS FOR LETTER AND CHARACTER RECOGNITION

  D.А. Bezuglov , М.S. Mishchenko , S.E. Mishchenko

  134-144

  2025-07-24

  Abstract ▼

  The accuracy of text image recognition remains limited in practice. This is due to the fact that the alphabet of symbols can include lowercase and uppercase letters with a similar font, as well as composite characters formed from several simpler characters. To solve this problem, the character recognition system is supplemented with semantic or structural analysis systems, which significantly complicates the information system for text recognition. Currently, convolutional neural networks are widely used for recognizing single characters, for which a database with images of recognized characters is used for training. The paper proposes an algorithm characterized in that the image of a single character for a training sample includes fragments of characters that can be located in a line in close proximity to the recognized character.  This allows you to expand the set of images for training and additionally include information in the image about the placement of the symbol in the string, its relative size and whether this symbol is composite. The formation of images for the training sample simulates the process of segmentation of a symbol by brightness, which is usually used when selecting a symbol for further recognition.
  At the same time, the size of the symbol is estimated, it is supplemented with images of neighboring symbols, and then the size of the area, the image that will be placed in the training sample, is estimated. The resulting image is scaled and cropped in such a way that images of a given size are received at the input of the neural network. In the work, to recognize the alphabet of symbols, including uppercase and lowercase characters of the Russian and English alphabets, numbers, symbols and punctuation marks, it is proposed to use a variety of convolutional neural networks, each of which is trained to recognize one character. The symbol is selected by comparing the responses of all neural networks and selecting the maximum response. The proposed algorithm for training data preparation is compared with a well-known algorithm based on the use of images of single characters. It is established that the proposed algorithm for preparing data for training provides an increase in the accuracy of recognizing the alphabet of 138 characters by more than two times.

• SEMI-MARKOV MODEL OF TELECOMMUNICATION NETWORK WITH DYNAMIC CONTROL

  D.A. Mishchenko, А.А. L’vov, А. А. Nikiforov, Alalvan Amin Raad Jihad, M.S. Svetlov

  2021-12-24

  Abstract ▼

  The paper proposes a semi-Markov model of telecommunication network. The variant of dynamic
  traffic control of queuing system as a special case of telecommunication network is considered.
  The main purpose of control is to minimize the average cost per unit of time to service the
  incoming flow of information (packets). This takes into account the different bandwidth of the
  channels, the processing speed of information in the channel and the information capacity of the
  buffers. The approach to the organization of dynamic control taking into account noise immunity
  (information reliability) and information security is discussed. The problem of dynamic control of
  a telecommunication network is considered on the example of a simple single-channel structure of
  the “point-to-point” type, which is modeled as a linear unidirectional Markov chain. The parameters
  of the service tariff, the cost of the fine for refusal of service were introduced. The analysis
  allows us to make the following remarks that the distribution of the input information flow of
  packets is Poisson, the law of distribution of the length of packets and the speed of their arrival is
  exponential, which together characterizes the Markov process. However, there are concurrent
  service delays relative to the timing of service requests, including buffer overflow delays. The proposed
  semi-Markov model of a telecommunications network can be used for more complex network
  structures. In particular, for a telecommunication network, consisting not only of one singlechannel
  information transmission system (single-channel queuing system), but representing a set
  of several systems, that is, for multichannel telecommunication networks.

• STUDY OF RETRANSMISSION SCHEMES IN WIRELESS SENSOR NETWORKS

  Alalvan Amin Raad Jihad, Shammari Najim Abed Mandila, D.A. Mishchenko, А. А. L’vov, M.S. Svetlov

  2021-12-24

  Abstract ▼

  Wireless sensor networks (WSNs) are being actively implemented in various systems for remote
  observation and monitoring of distributed objects. WSNs have a number of undoubted advantages:
  flexibility, efficiency, relative cheapness, and the possibility of rapid deployment. However,
  the exchange of information and data is carried out in the WSN using wireless communication
  channels, which are subject to inevitable interference and noise, which leads to transmission errors and even to the loss of transmitted data packets. Another challenge, not fully resolved, is the
  uneven distribution of consumed energy within the WSN in the face of stringent requirements for
  energy sources. Currently, there are two most widely used retransmission schemes in the loss of
  transmitted data, namely, hop-by-hop and end-to-end. Most of the well-known studies devoted to
  the issues of reliable data transmission in WSN using these schemes have been carried out experimentally.
  In addition, there are still no analytical methods for evaluating various reliable
  transport solutions, which complicates the analysis of the proposed WSN. Therefore, the aim of the
  proposed work is the development of analytical methods and algorithms for studying the operating
  characteristics of the signal relaying circuits in the WSN. Analytical methods are proposed for
  evaluating retransmission schemes in the WSN, based on a relatively new theoretical basis - the
  network calculus for packet-switched networks, which is a tool for determining the size of the network.
  First, traffic, service and energy cost models are introduced. Based on these models and
  network calculations, the maximum packet transmission delay and energy efficiency of the two
  main types of retransmission schemes: hop-by-hop and end-to-end retransmission, are analytically
  estimated. According to the results of the experiment, the maximum latency and the maximum
  power consumption of these two schemes are compared in several scenarios. In addition, the analytically
  calculated maximum delay is compared with the simulation results. With the proposed
  method, a suitable retransmission scheme can be selected based on the various requirements and
  constraints to be set.