Enhancement of User-Level QoS in Audio-Video IP Transmission by Utilizing the Mutually Compensatory Property

This paper studies a method of utilizing the cross-modal influences of audio and video, which is referred to as the mutually compensatory property here, in order to enhance user-level QoS (i.e., perceptual QoS) in IP networks. For that purpose, we clarify how the mutually compensatory property depends on the content type. We conducted an experiment on audio-video transmission over a simple IP network with ten contents which are classified into five types. In the experiment, we transmitted CBR audio-video streams with various values of guaranteed bandwidth for audio and that for video under the condition that the total guaranteed bandwidth of the two streams is constant. Then, we assess user-level QoS and perform QoS mapping between user-level and application-level by multiple regression analysis. Using the obtained regression lines, we examine the dependence of the property on the content type. As a result, we confirm that the mutual compensation relationship between audio and video can be classified into three kinds. Furthermore, we investigate how we should allocate the bandwidth between audio and video according to the content type in order to maximize the user-level QoS.