Augmenting Speech Quality Estimation in Software-Defined Networking Using Machine Learning Algorithms

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27240%2F21%3A10247444" target="_blank" >RIV/61989100:27240/21:10247444 - isvavai.cz</a>

Result on the web

<a href="https://www.mdpi.com/1424-8220/21/10/3477" target="_blank" >https://www.mdpi.com/1424-8220/21/10/3477</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.3390/s21103477" target="_blank" >10.3390/s21103477</a>

Result language

angličtina

Original language name

Augmenting Speech Quality Estimation in Software-Defined Networking Using Machine Learning Algorithms

Original language description

With the increased number of Software-Defined Networking (SDN) installations, the data centers of large service providers are becoming more and more agile in terms of network performance efficiency and flexibility. While SDN is an active and obvious trend in a modern data center design, the implications and possibilities it carries for effective and efficient network management are not yet fully explored and utilized. With most of the modern Internet traffic consisting of multimedia services and media-rich content sharing, the quality of multimedia communications is at the center of attention of many companies and research groups. Since SDN-enabled switches have an inherent feature of monitoring the flow statistics in terms of packets and bytes transmitted/lost, these devices can be utilized to monitor the essential statistics of the multimedia communications, allowing the provider to act in case of network failing to deliver the required service quality. The internal packet processing in the SDN switch enables the SDN controller to fetch the statistical information of the particular packet flow using the PacketIn and Multipart messages. This information, if preprocessed properly, can be used to estimate higher layer interpretation of the link quality and thus allowing to relate the provided quality of service (QoS) to the quality of user experience (QoE). This article discusses the experimental setup that can be used to estimate the quality of speech communication based on the information provided by the SDN controller. To achieve higher accuracy of the result, latency characteristics are added based on the exploiting of the dummy packet injection into the packet stream and/or RTCP packet analysis. The results of the experiment show that this innovative approach calculates the statistics of each individual RTP stream, and thus, we obtain a method for dynamic measurement of speech quality, where when quality decreases, it is possible to respond quickly by changing routing at the network level for each individual call. To improve the quality of call measurements, a Convolutional Neural Network (CNN) was also implemented. This model is based on two standard approaches to measuring the speech quality: PESQ and E-model. However, unlike PESQ/POLQA, the CNN-based model can take delay into account, and unlike the E-model, the resulting accuracy is much higher.

Czech name

—

Czech description

—

Type

J_imp - Article in a specialist periodical, which is included in the Web of Science database

CEP classification

—

OECD FORD branch

20203 - Telecommunications

Project

<a href="/en/project/LM2018140" target="_blank" >LM2018140: e-Infrastructure CZ</a><br>

Continuities

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>S - Specificky vyzkum na vysokych skolach

Publication year

2021

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Name of the periodical

Sensors

ISSN

1424-3210

e-ISSN

—

Volume of the periodical

—

Issue of the periodical within the volume

neuvedeno

Country of publishing house

CH - SWITZERLAND

Number of pages

14

Pages from-to

—

UT code for WoS article

000662513600001

EID of the result in the Scopus database

—