Design and implementation of a large scale audio-conferencing system
Paquay, Michaël
Promoteur(s) :
Mathy, Laurent
Date de soutenance : 7-sep-2017/8-sep-2017 • URL permanente : http://hdl.handle.net/2268.2/4291
Détails
Titre : | Design and implementation of a large scale audio-conferencing system |
Titre traduit : | [fr] Conception et implémentation d'un système de télé-conférence audio à grande échelle |
Auteur : | Paquay, Michaël ![]() |
Date de soutenance : | 7-sep-2017/8-sep-2017 |
Promoteur(s) : | Mathy, Laurent ![]() |
Membre(s) du jury : | Leduc, Guy ![]() Donnet, Benoît ![]() |
Langue : | Anglais |
Nombre de pages : | 72 |
Mots-clés : | [en] Audio conferencing [en] VoIP [en] Application Level Multicast [en] Next speaker prediction [en] Java [en] Java Sound API [en] Opus audio codec |
Discipline(s) : | Ingénierie, informatique & technologie > Sciences informatiques |
Institution(s) : | Université de Liège, Liège, Belgique |
Diplôme : | Master en sciences informatiques, à finalité approfondie |
Faculté : | Mémoires de la Faculté des Sciences appliquées |
Résumé
[en] Due mainly to the strict latency requirements of voice communication, variable network propagation delay and possibility of packet loss in audio transmission over Internet, building an audio-conferencing application represents a significant challenge in order to ensure good call quality, even more in a large scale context with hundreds of participants. However, some tests showed that participant’s tolerance about transmission delays is related to their active participation in conversation. This Master’s thesis presents the conception and implementation of a large scale audio-conferencing application, based on an activity prediction model and an application-level multicast protocol, called ALNAC, in order to minimize the perceived latency for all of the audio-conference participants by prioritizing the routing of audio packets directly to a subset of participants which are likely to be the most sensitive to latency. Moreover, a particular attention was given to the audio capture, playback and transmission. The latter is optimized by using an efficient audio codec, which is able to perform a high quality compression of voice and to mask at best the effects of jitter and packet loss. The result of this work is a basic and functional multithreaded Java application allowing to create or join a conversation and to take part in it, using the Java Sound API for handling audio operations and the Concentus library, which is a Java implementation of the Opus audio codec. This work first describes ALNAC and TBCP, an overlay spanning tree building control protocol used for the application-level multicast, before the use cases of the application and the software architecture. It then follows some explanations about the technologies, implementation details of the main operations and an overview of the graphical user interface. Finally, the thesis exposes a brief performance analysis and gives a conclusion about the work accomplished as well as some suggestions of improvement.
Fichier(s)
Document(s)
Annexe(s)
![File](/static/img/item/file.png)
![Accès privé Access](/static/img/item/file/pdf_priv.png)
Description: -
Taille: 553.88 kB
Format: Unknown
Citer ce mémoire
L'Université de Liège ne garantit pas la qualité scientifique de ces travaux d'étudiants ni l'exactitude de l'ensemble des informations qu'ils contiennent.