Home // ICAS 2018, The Fourteenth International Conference on Autonomic and Autonomous Systems // View article

Using Neural Networks to Evaluate the Intelligibility of the Speech in Voice over Internet Protocol

Authors:
Angel Garabitov
Aleksandar Tsenov

Keywords: speach; VoIP; intelligibility; neural network.

Abstract:
An important and unresolved problem is the automatic quantification of voice quality. Many parameters affect the voice quality, but only subjective assessments are crucial. The article makes a proposal to create a solution for the prediction of subjective voice quality assessment. A neural network is used to evaluate the quality of Voice over IP (VoIP) solutions. Computational models of similarity, Musical Instrument Digital Interface (MIDI) analysis and Speech Processing Tools are used to assess intelligibility. Statistics techniques and algorithms are used for building neural networks. The end result is an analysis and an automated hypothesis about sound quality in the Mean Opinion Score (MOS) scale.

Pages: 52 to 55

Copyright: Copyright (c) IARIA, 2018

Publication date: May 20, 2018

Published in: conference

ISSN: 2308-3913

ISBN: 978-1-61208-634-7

Location: Nice, France

Dates: from May 20, 2018 to May 24, 2018