Nicola Cranley

User perception of adapting video quality

In general, video quality adaptation and video quality evaluation are distinct activities. Most adaptive delivery mechanisms for streaming multimedia content do not explicitly consider user-perceived quality when making adaptation decisions. Equally, video quality evaluation techniques are not designed to evaluate instantaneous quality where the quality is changing over time. We propose that an Optimal Adaptation Trajectory (OAT) through the set of possible encoding exists, and that it indicates how to adapt encoding quality in response to changes in network conditions in order to maximize user-perceived quality. The subjective and objective tests carried out to find such trajectories for a number of different MPEG-4 video clips are described. Experimental subjective testing results are presented that demonstrate the dynamic nature of user perception with adapting multimedia. The results demonstrate that adaptation using the OAT out-performs conventional adaptation strategies in which only a single aspect of the video quality is adapted. In contrast, the OAT provides a mechanism to adapt multiple aspects of the video quality thereby giving better user-perceived quality in both the short and long term.

User-perceived quality-aware adaptive delivery of MPEG-4 content

Many adaptive delivery mechanisms have been devised for streaming multimedia over best-effort IP networks. Most of these adaptive schemes do not consider the users perception of quality when making adaptations. We propose that an optimum adaptation trajectory exists which indicates how encoding quality should be adapted (upgraded/downgraded) with respect to user perceived quality in response to network conditions. This optimum adaptation trajectory can be used with any transmission adaptation policy. We describe a system architecture that uses knowledge of user perceived quality to make adaptation decisions and give an example of how this knowledge can be used to complement the sender-based adaptation algorithm, LDA.

Performance evaluation of video streaming with background traffic over IEEE 802.11 WLAN networks

There is an increasing demand for multimedia streaming applications over WLAN networks. MPEG-4 and H.264 are compression standards targeted at high-quality streamed multimedia services over wireless best-effort IP networks. However, the dynamic nature of wireless networks in terms of fluctuating bandwidth and time-varying delays makes it difficult to provide good quality streaming under such constraints. Multimedia streaming applications are a demanding and challenging service to deliver over wireless networks. There is a trade-off between the capacity of the wireless network and the quality of the multimedia streaming application. In this paper we investigate the effect the background traffic load has on unicast streaming video sessions. We show that above a certain load value, the video streaming session is slowly starved of bandwidth. The load value at which this occurs depends on the characteristics of the background traffic load in terms of packet rates and the number of sources contributing to the load.

Optimum adaptation trajectories for streamed multimedia

Most adaptive delivery mechanisms for streaming multimedia content do not explicitly consider user-perceived quality when making adaptations. We propose that an optimal adaptation trajectory through the set of possible encodings exists and that it indicates how to adapt encoding quality in response to changes in network conditions to maximize user-perceived quality. Such an optimum adaptation trajectory can be used with any transmission adaptation policy. We describe the subjective tests we carried out to find such trajectories for a number of different MPEG-4 video clips and indicate how this knowledge could be used in the operation of a practical system.

The Effects Of background Traffic on the End-To-End Delay for Video Streaming Applications over IEEE 802.11B WLAN Networks

The bursty nature of video streaming applications is due to the frame-based structure of video and this has an important impact on the resource requirements of the WLAN, affecting its ability to provide quality of service (QoS) particularly under heavily loaded conditions. In this paper we analyse this bursty behaviour in depth. We show how each video frame is queued at the AP causing the packet delay to vary in a sawtooth manner that is related to the frame rate, the number of packets per video frame, and the packet size. We infer the maximum background traffic load that can be supported so that it does not negatively impact on the video streaming application. We demonstrate that there is a critical threshold load value above which the AP can no longer reliably support the video stream and compare it to the threshold load values calculated through analysis. Using this knowledge, the AP can employ resource allocation mechanisms to regulate the incoming traffic to the AP transmission queue so that QoS can be provided for streaming applications

Adaptive quality of service for streamed MPEG-4 over the Internet

Currently multimedia is either downloaded before viewing, or streamed over a network. However, the problem of streaming real-time or near real-time applications with a specified quality of service (QoS) over the Internet is still unsolved. The Real-time Transport Protocol (RTP) can be used to facilitate streaming, but also has the potential to support QoS. By gathering network statistics during the session and defining different QoS levels, we propose to adapt multimedia streaming to a fluctuating network load and/or client requests, thereby providing adaptive QoS. We describe the simple server and client applications we have implemented to illustrate this adaptation process.

Video Frame Differentiation for Streamed Multimedia over Heavily Loaded IEEE 802.11e WLAN using TXOP

In this paper we perform an experimental investigation of using video frame differentiation in conjunction with the TXOP facility to enhance the transmission of parallel multimedia streaming sessions in IEEE 802.11e. The delay constraints associated with the audio and video streams that comprise a multimedia session pose the greatest challenge since real-time multimedia is particularly sensitive to delay as the packets require a strict bounded end-to-end delay. Video streaming applications are considered to be bursty. This burstiness is due to the frame rate of video, the intrinsic hierarchical structure of the constituent video frame types, and the different compression ratios for the different video frame types. The TXOP facility is particularly suited to efficiently deal with this burstiness since it can be used to reserve bandwidth for the duration of the packet burst associated with a packetised video frame. Through experimental investigation, we show that there is a significant performance improvement for video streaming applications under heavily loaded conditions by differentiating between the constituent video frame types. The results show that video frame differentiation reduces the mean loss rate by 12% and increases the mean PSNR by 13.1dB.

Performance analysis of network-level QoS with encoding configurations for unicast video streaming over IEEE 802.11 WLAN networks

Video streaming has a large impact on the resource requirements of the WLAN. However, there are many variables involved in video streaming, such as the video content being streamed, how the video is encoded and how it is sent. This makes the role of radio resource management extremely difficult. In this paper we investigate the effect that video encoding configurations has on the network resource requirements for unicast video streaming in a WLAN environment. We compare the network resource requirements of several content types encoded at various encoding configurations with varying I-frame frequencies, target encoding bit rates and hint track settings. We present two key findings. We show that by halving the hint track MTU values, the access requirements of the WLAN are increased by 20%. Furthermore, we show how the I-frame frequency of the encoded file relates to the resource requirements of the WLAN.

An Experimental Investigation of IEEE 802.11e TXOP Facility for Real-Time Video Streaming

Real-time multimedia streaming applications require a strict bounded end-to-end delay and are considered to be bursty as each video frame is typically transmitted as a burst of packets. In this paper we show how the distribution of video frame sizes can be used to efficiently dimension the IEEE 802.lie TXOP limit parameter to efficiently deal with this burstiness in order to enhance the transmission of real-time video streaming services. Through experimental investigation, we show that by using the mean video frame size to dimension the TXOP limit parameter, the transmission delay for the video frame is reduced by 67% under heavily loaded conditions. Other techniques investigated in this paper include applying the TXOP facility separately to each of the constituent I, P, and B video frame types.

Perceptual quality adaptation (PQA) algorithm for 3GP and multitracked MPEG-4 content over wireless IP networks

Many adaptive delivery mechanisms have been devised for streaming multimedia over best-effort IP networks, such as the Internet. Most of these adaptive schemes do not consider the users perception of quality when making adaptations. We describe a perceptual quality adaptation algorithm (PQA) and prototype system architecture that uses knowledge of user perceived quality to make adaptation decisions using an optimum adaptation trajectory. This optimum adaptation trajectory indicates how encoding quality should be adapted (upgraded/downgraded) with respect to user perceived quality in response to rapidly fluctuating network conditions. We present simulation results that demonstrate the behavior of a perceptual quality adaptation algorithm.