Analyse et modélisation de la qualité perçue des applications de visiophonie

Abstract : In a highly competitive environment, one of the key challenges for operators and providers of video telephony services is to ensure the highest quality of experience (QoE). There is a strong need for a measure that reflects users satisfaction and perception of these services. The audio-visual quality of a video call must be controlled to meet two main needs. The first concerns the planning of new technologies under development. The second is focused on the control of existing communications by assessing the quality of the services offered and evaluating them. Two approaches are used to evaluate audio-visual quality: subjective tests by collecting scores given by participants on quality scales, after viewing and listening to audiovisual sequences and objective metrics based on automatic audio/ video or audiovisual quality evaluation algorithms. Concerning telephony services, decades of research, standardization work and network exploitation, have allowed operators to master the automatic monitoring tools and to determine the representative metrics of voice quality. However, the metrics for measuring the audiovisual quality of a conversational services are not yet mature and not exploited by telecommunication operators. The present work focuses on finding representative metrics of the perception of the video telephony anc videoconferencing services quality. These objective metrics are calculated from the audio and video signals. Subjective tests are conducted to collect the judgment of service users on the perceived quality according to different levels of degradation. We studied the impact of network conditions (packet loss, jitter and desynchronization) on the QoE of a video call. The general principle is then to establish a correlation between the selected objective metrics and the perceived quality as expressed by the users. The results showed that new metric of overall audiovisual quality that take into account the temporal aspect of video are more powerful than image quality based metrics. On the other hand, the use of a machine learning approach represents a solution to generat a global quality prediction model from the degradation metrics (blur, pixelization, image freezing, ... )
Complete list of metadatas

Cited literature [140 references]  Display  Hide  Download
Contributor : Abes Star <>
Submitted on : Thursday, January 10, 2019 - 3:54:16 PM
Last modification on : Wednesday, January 30, 2019 - 11:48:53 AM
Long-term archiving on : Thursday, April 11, 2019 - 4:41:14 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01977199, version 1


Inès Saidi. Analyse et modélisation de la qualité perçue des applications de visiophonie. Traitement du signal et de l'image [eess.SP]. INSA de Rennes, 2018. Français. ⟨NNT : 2018ISAR0013⟩. ⟨tel-01977199⟩



Record views


Files downloads