Technology to Power the New Rules: Delivering on the Video Opportunity Promise

By P. Wilford, E. Six

Introduction

An increasing number of carriers have arrived at a crossroads in the evolution of their networks. As they consider adding the capability to deliver advanced video services to their customers, and in order to support new business models for the communications economy of the 21st century, there is a need to get more content over the last mile. While many wireline carriers are in a strong position to offer initial services to a segment of their customers, fiber will have to be extended deeper into the access plant to provide uniform services to all their customers. In fact, many service providers today are updating their access plant with fiber-to-the-home. Moreover, core network bandwidths will also need to be expanded to support the scalability of wireline and wireless services across their entire customer base. But as important as increasing network bandwidth, powerful processing elements in the network must be added to handle the rigorous demands of advanced video services. This article highlights several significant opportunities and issues relating to IPTV, and demonstrates particular solutions that Alcatel-Lucent has developed to address these needs.

There is no doubt that the number of IPTV subscribers will grow significantly. Some of this growth will come from enhancements of existing access networks that are currently not able to support video services. Other growth will be the result of advanced video services that attract new subscribers. Further, by addressing some fundamental issues in the delivery of video service, as will be described in this article, IPTV will be more attractive. Market forecasts predict a rise from an IPTV subscriber base of 13.5 million in 2007 to 72.6 million in 2011 (Figure 1). This represents a compound annual growth rate of 40%.



Corresponding revenue streams from IPTV will grow from $3.7 billion USD (€2.4 billion) in 2007 to $23.2 billion USD (€14.9 billion) in 2011.

In addition to providing IPTV over the wireline access network, delivery of video over the wireless cellular network to handsets has become significant in the past two years. This represents not only a change in required technologies and networks, but also a change in the type of content consumers want and will watch and pay for. Trends like viewing shorter clips, interactive viewing, larger screens and smaller mobile screens all create challenges for the video delivery network.

Recently, we have seen new network architectures associated with video distribution. One such approach, hybrid satellite and terrestrial wireless systems, offers the advantage of multicast broadband downstream capability (for "conventional" TV watching) while enabling more flexible unicast services as well. Again, these types of networks require innovative solutions to ensure the quality of delivery subscribers expect, particularly if they are paying for the service.

However, the good news for most carriers is that they will be able to leverage much of their existing networks and their current network elements to deliver a new generation of video services. Alcatel-Lucent is working on management systems and video processing elements that can be added to their current equipment. As carriers upgrade their system to increase bandwidth and allow the delivery of advanced video services, they will be in an increasingly better position to make incremental investments, allowing them to cost-effectively take advantage of the growing revenue opportunities.

Challenges to Overcome

As carriers explore the opportunities presented by new demands for video services over their networks, there are technical challenges that must be solved to fulfill the promise of the video opportunity. One of the most compelling issues revolves around how to ensure that the network delivers video services - and all the associated video features and services customers have come to expect from this category of services - so that latent demand can be converted into new revenue-generating opportunities.

It's in this area that Bell Labs has developed significant innovative breakthroughs. Video delivery for carriers has proven to be very challenging. Almost all the difficulties relate back to the fundamental fact that the industry has mixed two very successful technologies together to build video systems:

• MPEG video compression and
• IP transport

While these two technologies have been extremely successful in their own rights, they were never intended to be mixed. MPEG, for instance, is a very successful standard for video compression. It was created with a paradigm of supporting long video sequences (such as movies or TV programs) with well engineered broadcast channels and storage-intensive playback devices (like DVD players). The compression uses a lot of prediction to remove redundancy from the signal and has long-time constants to manage video buffers and other resources. These techniques have resulted in fantastic compression (better than 50 to 1) and picture quality like we are beginning to experience with ultra-HDTV displays and soon-to-come 3D TV, and now extending to multiple screens and many new types of players and recorders (like Blu-ray).

IP networking and transport is also a very successful technology. It allows for all kinds of very different services to be carried over the same network. IP packets travel across the network independent of each other and pack the higher-level services into small segments. This allows for greater flexibility and the ability to share resources of network equipment, and results in the rise of very low-cost transport and delivery networks.

But when these two technologies (developed independently of each other) are combined and integrated, four key problems emerge that must be addressed before new video services can be successfully brought to market:

• Reliable Delivery: Problem is packet delivery can have a profound effect on the picture quality perceived by the customer. Because IP packets arrive at non-predictable and irregular times (and at times out of order or not at all), it poses a serious challenge to an MPEG standard that is time-sensitive and dependent on highly predictive sequencing. The way the signal is received by the subscriber is not the way a traditional MPEG receiver was designed to expect it.
• Channel Change: Because of the careful way any MPEG stream is created, both in terms of prediction and buffer control, an instantaneous random switch to a different stream will cause multiple seconds of garbled video until things correct themselves. Solutions must be deployed to alleviate this problem.
• Advertisement Insertion: Because of the nature of the MPEG signal, switching from the main program stream to an advertisement often results in "wrong predictions" and other problems in the stream. These must be addressed properly to keep the subscriber from viewing severe artifacts for several seconds at each advertisement boundary. (See our article in this issue, entitled Hitting the Target with Television Advertising)
• Delivery Over Wireless Networks: To enhance the revenue generating capability of a carrier's network, there is a desire to send the same content over networks with much lower bandwidths. Therefore, there is a need to take MPEG content and reduce the bandwidth significantly. Solutions are required to be able to do this in the compressed domain.

Reliable delivery

To get a feeling for the challenges that must be overcome for these areas, please refer to Figures 2a-d. As will be seen, all four of these issues are related to the properties of the MPEG and IP technology and will share a common set of solutions.

We start with reliable delivery, which is described based on packet transmission shown in Figure 2a. Shown is a series of packets, each containing a successive portion of the video signal. The first and sixth packets are "anchor" frames. These anchor packets contain all the aspects of the video signal needed to display a high quality picture on the subscriber's screen. The packets in between are "predicted" packets. We show in Figure 2a a situation where an error is made in the transmission of packet number 4. This could be anything from a bit error, all the way up to a lost packet. The significant issue for IPTV systems is that, in this scenario, not only is the picture corrupted during the time of the "errored" packet, but also the picture stays corrupted or impaired for all subsequent packets. This is because the prediction is based on something that is itself corrupted. This problem situation persists until another anchor packet is received. In the figure, the red oval represents the time when the picture displayed on the screen has residual corruption. It could last for several seconds. To further illustrate the situation, the figure shows what the TV picture would look like. Note the corruption persisting from the point of the error all the way down until the next anchor frame.



One way to solve the problem of reliable delivery is to ensure that all packets arrive to a customer in the way an MPEG system would like them to arrive. This is an approach that has been tried by early IPTV systems that have been deployed. But it is a very expensive and unrealistic approach, because it is trying to convert the IP network into something that it is inherently not.

Channel change

The root cause of very slow channel changes in IPTV systems is caused by the same fundamental aspects that led to packet transmission problems - specifically the predictive nature of MPEG. Figure 2b shows the same type of packet transmission and prediction. Let's suppose the subscriber is watching channel A. Then at the packet number 3 timeframe, the subscriber switches to channel B. Since channel B is transmitting "predicted" packets, the resulting picture displayed on the TV screen will be corrupted, because the predicted frames are from the old channel. This situation will again persist until a new anchor frame is received. This situation can result in very slow channel change times; it can take several seconds. With new advanced codecs, such as H.264, this situation is getting worse because there is a longer time between anchor frames.



To address the problem of slow channel change (and also improve the reliable delivery problem), one could make the anchor frames appear a lot more frequently in the bit stream - but this would create higher bandwidth requirements or result in the reduction of overall picture quality.

Advertisement insertion

Local advertisement insertion is an important revenue generating capability of a service provider offering video services. It refers to the capability of network operators to be able to locally (small town size geography) insert customized advertisements for that particular locality. However, as we show in Figure 2c, the same nature of MPEG makes ad insertion challenging. Suppose the service provider wants to insert a local advertisement at the time of packet number 3. The original bit stream was very carefully set up using the predictive nature of MPEG. In addition, there are many key parameters in the bit-stream - like buffer levels, timing information and global parameters. The advertisement being spliced into the signal will have parameters that are different. This change of parameters that results when an ad is inserted can lead to a problematic and noncompliant bit-stream, which will result in a corrupted picture, and may even cause reset problems in IPTV set-top boxes (STBs).



The problem of local ad insertion could also be addressed by coding the video in a prescribed manner but, again, it would be at the expense of picture quality.

Transcoding - Delivery over wireless networks

The last issue, transcoding and transrating, presents a slightly different and difficult challenge. Addressing and solving this challenge, however, is significant because if it can be done in a cost-effective and manageable manner - in the IPTV Network - it will open up new revenue streams for service providers as they now have the capability to re-purpose content to many more types of subscribers over many more types of access networks.

A common scenario is as follows. Video to be delivered on an IPTV network will have been prepared and encoded at a certain bit rate and with a certain format, chosen to work with the characteristics of the access network and type of end-user device. A typical example would be 3 Mbps using MPEG2, assuming a DSL network with the most common type of STBs. However, service providers would like to have the ability to take these video streams and deliver them to subscribers using networks that have significantly lower bandwidth and with end-devices that support a different syntax. In Figure 2d we show a "transrating" from 3 Mbps down to 500 Kbps and a "transcoding" from MPEG2 to H.264. These rates would be common of what's required for video on a wireless network.



The need to transcode and transrate the video could be solved by decoding the video and then re-coding it at a different rate and syntax. But this would be very expensive. Alternatively, multiple bit-streams with different rates and formats could be provided at the access point and a selection made at that point. But this is again expensive and requires more bandwidth and management.

Alcatel-Lucent's Approach

Alcatel-Lucent has developed a unified solution to cost-effectively address all four issues. It involves placing a video processor and server with buffering into the existing network elements. This technology is shared by all the subscribers, and thus it is extremely cost-effective. Further, since the solution is integrated into the existing network elements, it can be managed as part of the existing infrastructure. And for legacy systems, we have architected the solution so that it can be placed in a stand-alone video-processing server.

The key to our solution revolves around placing a powerful video-processing server into our access and edge router equipment. This video-processing server is made of three major components and is shown in Figure 3:

• Video packet processors
• Video content processors
• Video packet buffers

The buffer stores a small segment of each video stream being delivered by the network. This would typically be 100 to 500 TV channels and five to 10 seconds of each channel. The video packet processor manages the video stored in the buffer.

In order to improve Reliable Delivery in this environment, when an error in the bit-stream is received by the STB - which would otherwise result in severe picture impairment - the STB signals the network to resend the packet that has an error. The STB then replaces the bad packet with the good re-sent packet before the picture is decoded and displayed on the TV. There is a one or two second buffer in the STB, which allows time for this technique to work. Thus there are no visible impairments.

Fast Channel Change uses the same basic network functionality. When a subscriber clicks the remote to change channels, a signal is sent to the network to send the new selected TV channel. By using Alcatel-Lucent's solution, the channel is sent to the subscriber - not from an arbitrary point, which would result in several seconds of delay before a good picture would appear on the TV - but from an appropriate point in the buffer so that the picture can immediately be properly seen. The video processor will then "speed up" or "slow down," by either adding more compression or increasing the transmission rate to bring the buffer back to the normal operating position. This process takes between five to 10 seconds. The buffer re-synchronization occurs without any visual perception by the customer.

Seamless Ad Insertions utilize the same network video-processing capability. Our approach allows for the processing of the main TV program and the advertisement to be integrated into the signal so that when the splice is made, the bit-stream is fully compliant, allowing the STB to decode it and display the signal on the screen without any artifacts. The video-processing element we have developed has the capability to match anchor frames with splice points and perform all the required signal processing and manipulation to make the splice seamless.

One of the most powerful aspects of IPTV systems is the ability to deliver truly targeted or personal advertisements to individual subscribers. The techniques described here developed by Bell Labs allow for the seamless delivery of personal advertisements to every individual subscriber without any modifications of existing STBs.

The Transcoding and Transrating function requires extensive processing of the video signal. This is where the role of the video processor is critical. The approach we have taken is to process the video while it is still in the compressed domain and reduce the bandwidth and change the coding syntax to adapt to the available access bandwidth and the type of client device that is receiving the signal. We use a variety of techniques to make this happen, including:

• Motion vector sharing,
• Re-quantizing coefficients and
• Reducing the bit rate of select codes.

All three of these techniques can be executed much more efficiently (in terms of required processing power) within the video-processing server than using conventional methods.

Alcatel-Lucent is implementing a product and architecture strategy for adding this video-processing capability into our network equipment. We have developed a common set of algorithms and protocols that all of our products will implement, thus ensuring that when customers deploy large-scale systems, these key aspects of video delivery will work very well.

Considerations for Implementing Video Strategies

One of the most important issues associated with deploying technologies that deliver high-quality video and video features customers are demanding is to ensure that the technology implementation is coordinated across all the network elements in the video-delivery system. Fast channel change and reliable delivery require signaling between the STB and the access and edge router equipment. Seamless ad insertion is most effectively accomplished when the routers are coordinated with the access equipment. Transcoding and transrating requirements require knowledge of access bandwidth and performance characteristics, as well as knowledge of the type of end-user devices. And to optimize performance (scalability), to minimize expense and to be able to manage the video subsystem as effectively as possible, an integrated view of the network must be considered.

That said, there are two barriers to deploying this new technology. The first is that there are large-scale existing legacy systems that service providers may want to use for IPTV offerings. The second is that new technology must be deployed in multiple places in the network simultaneously for this new capability to work.

To break through these barriers, Alcatel-Lucent has done two things:

• We have developed this new technology in a highly modular and flexible manner that enables its incorporation into all of our network equipment. This includes the ability to enable customers with legacy equipment to incrementally upgrade their network in a step-by-step fashion until all the new capability is incorporated in the network elements.
• We have tested our common algorithms and protocols in multiple systems test labs. In fact, our different product development teams used common core assets to speed our time-to-market and make testing performance in the field much quicker.

Conclusion

IPTV systems are still at the beginning of their deployment potential. They offer the promise of tremendous value to the end user - blended services, complete interactivity, simplified home networking as well has high-quality, big-screen, entertainment TV. The technology Alcatel-Lucent has developed ensures that nothing detracts from the user experience of watching TV in any format on any device.

Paul Wilford is Director, Video and Mobile Packet Networking Research Department, Alcatel-Lucent Bell Labs, Murray Hill, NJ, USA. Paul is a Bell Labs Fellow.

Erwin Six is Department Head, Video Technology, Alcatel-Lucent Bell Labs, Antwerp, Belgium.

To contact the authors or request additional information, please send e-mail to enrich.editor@alcatel-lucent.com.

Associated Items

Special Article (BLTJ)*: Analysis and Realization of IPTV Service Quality

Special Article (BLTJ)*: Monitoring Infrastructure for Converged Networks and Services

Special Article (BLTJ)*: IP-Driven Access-Independent Resource Management in Converged Access Networks