9.5 Conclusions
The capabilities of 3G networks and the suitability of different radio bearer configurations for supporting real-time video and voice communications were given in Section 9.2. A bearer configuration with a spreading factor of 8 is clearly unsuitable for video applications due to high channel-induced errors and high transmission-power requirements. The lower bound of Eb/No that could support acceptable video quality is found to be 8.5–10 dB for a spreading factor of 8 with a 1/3 rate convolutional code channel in vehicular and pedestrian environments. This is not achievable in a multi-user system, unless very sophisticated diversity techniques and interference-suppression techniques are employed. The minimum Eb/No requirement was reduced down to 6.3–8 dB for spreading factor 32 and 16 channels. Fast power control can lower the minimum Eb/No requirements further by 0.8–3 dB in tested operating environments. Perceptual quality measurements show that spreading factor 32 outperforms others for video transmission over similar operating conditions. However, this limits the operating source rate to 64 and 97 kbps with 1/3 and 1/2 rate channel coding, respectively. System-level performances demonstrate the necessity of advanced antenna techniques, adaptive resource allocation, and combined source-and channel-quality enhancement techniques in achieving 95% user satisfaction for real-time video services even for source rates as low as 64 kbps.
In addition to the investigation of capabilities of 3G networks for the transmission of video and voice, perceptual video quality enhancement methods that exploit the adaptive source and network parameters for varying transmission conditions were explored in the work carried out in this chapter. The most efficient way to use limited radio resources is the deployment of class- or priority-based channel allocation. In this way, the most important information is transmitted over a high-priority channel, while a low-priority channel is used to transport low-priority data. The syntax of the video format and the video representation can be used in information prioritization. Data partition formatting and object-based video coding provide simple methods of data separation in multi-priority transmission. However, a perceptual quality estimation-based data separation algorithm provides an adaptive and flexible information prioritization mechanism which results in optimal video performance. The priority of the transmission channel can in fact be based on the channel protection, in terms of channel coding rate, modulation, spreading factor, and transmission-power allocation. A combination of these channel parameters provides optimal prioritized transmission.
In addition, link adaptation, where source and network parameters are adjusted according to the time-varying channel condition at the receiver, is necessary in maintaining the perceptual video quality received by the end user. As discussed in Section 9.4, link adaptation can be designed either to enhance the received video quality for a given network configuration or to enhance the system performance in terms of system capacity or coverage, while guaranteeing end users' QoS requirements. Either way provides improved performances compared to non-adaptive transmission schemes. Two link-adaptation algorithms are implemented. One is designed to maximize the received video quality, and is named “quality-based adaptation scheme”. The other is designed to maximize the system throughput while guaranteeing a required video quality and is called “throughput-based adaptation scheme”. Source- and network-parameter adaptation is conducted according to the estimated instantaneous channel quality, based on measured channel BLER, RSS, and the first-order statistics of RSS. Algorithm performances are demonstrated for video telephony over an EGPRS network. The adaptive schemes showed better performances than those of non-adaptive schemes. When offered similar traffic load and channel environments, the quality-based adaptation scheme outperforms the throughput-based scheme. Furthermore, the investigation shows that the proposed algorithms are robust against feedback delay, noisy feedback, and burst channel errors.
The above-described link-adaptation algorithms are not suitable for video streaming applications as they require separate encoding of video sequences for each user. To overcome this problem, the video sequence is encoded at different output bit rates, which are stored in a buffer at the server. Link adaptation is performed by switching between pre-encoded streams according to the instantaneous channel condition. When switching is performed between streams, there is a mismatch between the encoder and decoder in terms of the reference frames used to predict future frames. This can potentially lead to drift. However, the proposed scheme uses an appropriate selection of intra-coded blocks (AIR technique) in video frames to limit the drift effects, and prevents this becoming a detectable problem. Experiments performed using EGPRS channel models show significant quality improvements for the adaptive scheme compared with the results for fixed modulation coding schemes.
Spreading gain provides the key variable in determining user data rates and associated channel quality in CDMA-based communication systems. Therefore, in addition to the channel coding schemes, adaptive spreading gain can be used to exploit time-varying channels in CDMA systems. Link-adaptation techniques, based on adaptive spreading gain control, are proposed and analyzed for real-time video communications in UMTS networks. Source rate and spreading code levels were varied depending on the state of the transmission channel. Adaptation is based on the actual channel signal-to-interference ratio, which is calculated at every TTI. The transmission power is kept at a constant level and the transmission bit energy is adjusted according to the selected spreading code level. This ensures that the interference power experienced by other users does not affect the adaptive spreading gain control techniques. The conducted experiments show 2–3 dB frame PSNR improvement compared to non-adaptive schemes. Further performance improvements are achieved by the combined application of adaptive spreading gain control and the unequal error-protection scheme.
A joint system-link UMTS simulator, which combines a system-level simulator and the developed link-level simulator, is designed for the analysis of video performances in a multiuser downlink UMTS system. Video performances are investigated for the transmission of fully error resilience-enabled MPEG-4-coded video under constraints on the base station total transmit power (power budget) and the number of codes available per carrier (code budget). Video performance is measured in terms of the average frame PSNR of received video by individual users. System performances are shown in terms of the mean video quality, which is the average video quality received by users, and the number of satisfied users in the system.
Three different resource-allocation schemes are implemented. Scheme 1 is a non-adaptive scheme, and it allocates equal transmit power for video users who require the same QoS. Scheme 2 adapts the allocated transmit power for each user according to their individual received channel quality. Scheme 3 combines an adaptive spreading gain control scheme with Scheme 2 to maximize the video quality received by individual users in the system. Experiments conducted over the simulated system show about 2–3 dB quality improvement in terms of average PSNR with adaptive power allocation (Scheme 2) compared to fixed resource allocation. Further performance enhancement, in terms of average quality and user satisfaction, is achieved with Scheme 3.
Finally, an enhanced user-centric CAC scheme in the downlink direction of a UMTS system was presented and evaluated using an implemented and validated WCDMA UMTS UTRAN system-level simulator. The scheme can be used to achieve the expected end-to-end QoS for an Enhanced-service-class user, providing lower blocking probability.