| Abstract [eng] |
In the times of informational society the importance of crucial information is the main factor for development of telecommunication sector. The problems to be solved are no longer only the transportation of info anywhere at any instance but also the minimal price to be paid for it. For last decade the technical solution to satisfy these objectives was moving towards IP network. The presenting qualification work is to study the voice transfer on IP network. First chapter is dedicated to realization of telephony over IP networks. The network parameters with most significance for VoIP quality are analyzed here. In the second chapter artificial neuron networks are presented as one of the possible VoIP Quality of Service clusterizators. An overview of their structures and modules as well as their advantages over other methods solving classification problems is made here. The model of artificial neuron network designed to analyze the QoS of voice transmission over IP network is given in chapter three. The model is based over network of perceptrons with two hidden layers. In the practical part of the work the designing of VoIP QoS forecasting artificial neuron network is made on NeuroSolutions for Excel program packet. The model is realized on network between Vilnius and Kaunas cities over which VoIP transmission is taking place. Using data from the real-world parameter measurements the research of QoS is made. Analysis of the results is made at the end. Results show that artificial neuron network is suitable for evaluation of QoS of VoIP. It is worthy to use this kind of neuron clusterizator like evaluator for QoS of VoIP gaining more control over the network by forecasting or notice at right time the fall of Quality of Service. Qualification work contains 40 pages including 27 pictures, 7 tables, 2 appendixes and 20 literature sources. |
