Please use this identifier to cite or link to this item: http://repositorio.ufla.br/jspui/handle/1/15468
Title: Impacto do desvanecimento no canal de radiofrequência na qualidade do sinal de voz
Other Titles: Impact of fading on the radio frequency channel in the voice signal quality
Authors: Rodríguez, Demóstenes Zegarra
Correia, Luiz Henrique Andrade
Silva, Bruno de Abreu
Malheiros, Neumar Costa
Conceição, Gabriel Fernando Pivaro Leite da
Keywords: Som – Registro e reprodução – Qualidade
Voz – Qualidade – Avaliação
Voz sobre IP (VoIP) – Degradação
Redes de computadores – Métricas
Sound - Recording and reproducing – Quality
Voice – Quality – Evaluation
Voice over IP (VoIP) – Degradation
Computer networks – Metrics
P.862
P.563
MOS
Issue Date: 4-Oct-2017
Publisher: Universidade Federal de Lavras
Citation: AFFONSO, E. T. F. Impacto do desvanecimento no canal de radiofrequência na qualidade do sinal de voz. 2017. 85 p. Dissertação (Mestrado em Engenharia de Sistemas e Automação)-Universidade Federal de Lavras, Lavras, 2017.
Abstract: Currently, voice over IP (VoIP) is one of the most popular communication services. However, there are many external factors that affect the global quality of a VoIP communication, the voice can suffer various types of degradations, especially when the transport is performed through wireless channels. Packet loss and fading are factors that significantly affect the voice signal, either in wired or wireless communication channels. Therefore, metrics to assess the quality of voice becomes relevant in order to ensure an adequate service and maximize user satisfaction. In this research, different degradation scenarios such as packet loss and fading models will be applied to voice files. Then, the voice quality index for each scenario will be determined by both ITU-T recommendations P.862 and P.563. Firstly, a new metric, called MOSp, was proposed based on the experimental tests, and using the P.862 as reference, with it is possible to predict the quality index of a given communication, using radio frequency parameters such as Doppler, the fading model and also including wired network parameters such as packet loss rate. Considering a MOS scale of 0 to 4.5, the metric had the largest error of 0.3354, representing a margin of error of only 7.4%. In addition to the development of the proposed MOSp metric, the training of an Artificial Neural Network (ANN) was also carried out, with the input of degradations and MOS values obtained through P.862, the training result reached a maximum error of 0.0843 and a percentage of only 1.87% of errer. Later, because of its non-intrusive characteristics, the P.563 model was used, but its results obtained were lower than expected, then, a new metric denominated P:563P MOS was developed considering the results obtained by P.862 as reference. The results of the new technique presented a Pearson Correlation Coefficient of 0.94 and a maximum error of 0.49 MOS. For a better representation of the scenarios presented in simulated environments and to identify if there is a natural degradation of a real transmission, experiments were carried out in indoor and outdoor environments with transmissions at different distances and modulations, these transmission were made with the use of Software Defined Radio (SDR) model implemented in Universal Software Radio Peripheral (USRP) devices. Thus, it was possible to obtain the best representation of the scenarios presented in simulated environments, to identify if there is a natural degradation of a real transmission. The experimental results validate that there is no loss of quality in a transmission that has not suffered degradation, and using the Binary Phase Shift Keying (BPSK) modulation, implemented in the USRP, there is a better efficiency of the voice transmission, reaching the maximum score of ITU-T P.862 quality scale.
URI: repositorio.ufla.br/jspui/handle/1/15468
Appears in Collections:Engenharia de Sistemas e automação (Dissertações)



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.