Subcategoría “Sentido de ayuda”

The optimization of the content is carried out by applying the QoS-driven model described in Chapter 5 at the receiver end in real time. The optimization of the network resources is dependent on finding the impact of QoS parameters on end-to-end quality for each type of video application. Through statistical analysis of ANOVA and PCA, the QoS parameters have been identified for each video application, hence enabling in the optimization of existing

7.3. QoS-driven Optimization of Content Provisioning and Network Resources

152

network resources. This is best explained by the flow diagram of the proposed QoS-driven adaptation scheme which is depicted in Fig. 7.1. From Fig. 7.1 the video application is first defined based on the content features (e.g. SM, GW or RM). Then from the video quality prediction model the process of optimization of content provisioning and network resources takes place based on either adapting the SBR or finding the impact of QoS parameters.

Figure 7.1. Flow diagram of the proposed QoS-driven scheme for optimization of content provisioning and network resources

QoS-driven prediction model

In order to establish the initial encoding sender bitrate for the three content types, the variables of FR and PER are fixed in Eq. (5.19). The FR was fixed at 10fps and PER as 0 assuming that there are no network losses. The SBR versus MOS curve is shown in Fig. 7.2 for the three content types (Fig. 5.6 is re-drawn to reflect SBR vs MOS as opposed to PSNR). The purpose of Fig. 7.2 is to show the maximum and minimum SBR achieve able highlighting the initial encoding requirement. Therefore, it shows the relationship of MOS with application QoS parameter of SBR. From Fig. 7.2 it is observed that there is a minimum

7.3. QoS-driven Optimization of Content Provisioning and Network Resources

153

sender bitrate for acceptable quality (MOS>3.5) for all content types. A MOS of 4 is considered “good” for streaming applications [161] where most users are satisfied. There is also a maximum sender bitrate for the three content types that gives maximum quality (MOS ~ 4.2). For example for the content category of SM, sender bitrate of 100kbps gives a maximum of 4.2. However, in RM higher sender bitrates are required for maximum quality i.e. > 500kb/s. From Fig. 7.2 it can be derived that when the sender bitrate drops below a certain threshold that depends on the type of video content, then the quality becomes ‘bad’. Moreover, the quality does not improve considerably for sender bitrates higher than a specific threshold, which also depends upon the spatial and temporal activity of the video clip. This is useful when applying adaptation to SBR as it defines the initial encoding bitrates for all content types.

Figure 7.2. MOS Vs Sender Bitrate for the three content types

The reference-free video quality models over wireless network are described in Chapter 5, Section 5.7 where video quality in terms of the MOS is predicted from a combination of encoder related parameters of Sender Bitrate (SBR), Frame Rate (FR) and WLAN access network related parameters of Packet Error Rate (PER) for three different video applications classified earlier as SM for video conferencing application, GW representing a typical video call and RM representative of video streaming (CT=0.5 in Eq. 5.19). The video codec for

0 100 200 300 400 500 600 2.6 2.8 3 3.2 3.4 3.6 3.8 4 4.2 Sender Bitrate(Kb/s) M O S SM GW RM Max MOS Min MOS

7.3. QoS-driven Optimization of Content Provisioning and Network Resources

154

these applications was MPEG4. The prediction model is obtained by nonlinear regression analysis of the QoS parameters both in the application and network level and is given as below in Eq. (5.19).

CDE  _{' PQL RQL}I JKL MNA OL1 (5.19)

For the purpose of demonstrating adaptation, no packet losses are assumed and hence PER = 0. The above Equation then reduced to Equation 7.1 below.

CDE  S _ Tgh _ Vln Eh (7.1)

The metric coefficients were re-fitted by non-linear regression of the prediction model with our training set (MOS values). The re-fitted metric coefficients α, β and γ along with the correlation coefficient showing the goodness of fit and RMSE for all three video applications over WLAN networks are given in Table 7.1.

Table 7.1 Re-fitted Metric coefficients

Coefficients SM GW RM S 2.797 2.273 -0.0228 T -0.0065 -0.0022 -0.0065 V 0.2498 0.3322 0.6582 R2 88.82% 89.19% 99.57% RMSE 0.1399 0.1354 0.0352

The model was trained with three video sequences of Akiyo, Foreman and Stefan in the three categories of SM, GW and RM, whereas the model is verified with three different video sequences of Suzie, Carphone and Football in the three corresponding content categories. MATLAB™ function nlintool has been used to carry out the nonlinear regression analysis. R2 indicates the goodness of fit of the fitted coefficients of the three models. See Chapter 5 for details.

The predicted video quality metrics are then used in the QoS-driven adaptation scheme to adapt the video sender bitrate as shown in Fig. 7.3.

7.3. QoS-driven Optimization of Content Provisioning and Network Resources

155

Figure 7.3. Block diagram of the proposed scheme

The basic model of our adaptation scheme, given in Fig. 7.3 consists of the following modules.

Content classifier: The content classifier classifies video sequences into three categories as SM, GW and RM using cluster analysis based on the spatial and temporal features of the video and hence determines the content type. . The category of SM represents video with low spatial and temporal movement, whereas, the category of RM represents video with high spatial and temporal movement. The third category of GW represents video of low/high spatial and high/low temporal movement. This is described in detail in Chapter 4.

Video Sender Bitrate Adaptor: This block adjusts the sender bitrate of the transmitted video as per the information received from the video quality prediction model given by Eq. (7.1) and according to Table 7.2.

Encoder: Lastly the adapted video clip is achieved according to the quality based on video content dynamics. Layered encoding is used for adapting the video streams to the content dynamics. The video clips are encoded in number of layers in a way that each layer increases the video quality of the video clip. Base layers are encoded at a very low rate to accommodate for different access networks (e.g. UMTS or WLAN). Additional layers are added to adapt the video stream according to the content type.

7.3. QoS-driven Optimization of Content Provisioning and Network Resources

156

Table 7.2 Experimental video scale assignment for SBR

Scale Value Average Sender Bitrate (Kb/s)

1 1 – 43 2 44 – 79 3 80 – 127 4 128 – 255 5 256 – 339 6 ≥340

The histogram of objective preference count (MOS values obtained from PSNR conversion [21] for all three content types is shown in Fig. 7.4. Fig. 7.4 shows that each content type has its unique patterns of adaptation preference supporting our adaptation scheme. The video clips are labelled as SM, GW and RM with low, medium and high complexity, respectively. A clear trend can be seen as the video content complexity increases outlining the adaptation preference as more complex videos need higher send bitrate for acceptable quality.

Figure 7.4. Histogram of the three content types

Sub-sections 7.3.2 show the application in content optimization and 7.3.3 show the application in network optimization.

1 2 3 4 5 0 5 10 15 20 25 30 35 40 MOS P re fe re n c e c o u n t SM GW RM

7.3. QoS-driven Optimization of Content Provisioning and Network Resources

157