TU Berlin

Quality and Usability Lab2018_08_27_Zhang

Inhalt des Dokuments

zur Navigation

Packet Loss Detection Training Based on a Speech Impairment Annotation GUI

LOCATION:  TEL, Auditorium 3 (20th floor), Ernst-Reuter-Platz 7, 10587 Berlin

Date/Time: 27.08.2018, 15:00-15:30  

SPEAKER: Yuezhu Zhang (TUB)



Much progress has been made in speech quality improvement in recent years. However, packet loss is still the major source of speech impairment in speech transmission. This kind of loss may cause by discarding packets in the IP networks due to congestion or by dropping packets at the terminal. The impact of packet loss on perceived speech quality depends on several factors, including loss pattern, codec type, packet size and loss locations[1]. Therefore, the goal of the bachelor thesis is to evaluate amount of speech file, to evaluate the quality of the speech file through internet transmission and do more to analyze the major reason for the impairment of speech file. With help of a specifically designed MATLAB App, which would be an Impairment Annotation GUI, it is possible to analyze a large amount of speech files in a short time, thus, some research effort has been placed on developing this app.

After that, with help of the MATLAB App we are able to train a prediction model that can detect missing packets in speech files, which were recorded from live calls. The trained detection model may could be used to estimate the quality of communication networks in further work.





Schnellnavigation zur Seite über Nummerneingabe