Mobile video QOE part II: Objective measurement

Objective measurement of video is performed using mathematical models and algorithm measuring the introduction of noise and the structural similarity of video objects.

There are several mathematical models such as PSNR (Peak Signal to Noise Ratio) and SSIM (Structural Similarity) for instance, that are traditionally used for these calculations. The complexity resides in the fact that a mathematical difference from one pixel to another, from one frame to another does not necessarily translate equally in the human eye.

PSNR is a measure that has a medium to low accuracy but is quite economic in computation. It represent possibly up to 10% of the CPU effort necessary to perform a transcoding operation. This means that although it provides a result that is not fully accurate, the model can be used to compute calculations as the file is being optimized. A vendor can use PSNR as a basis to provide a Mean Opinion Score (MOS) on the quality of a video file.

Video quality of experience measurement can be performed with full reference (FR), reduced reference (RR) or no reference (NR).

Full Reference

Full reference video measurement means that every pixel of a distorted video is compared to the original video. It implies that both original and optimized video have the same number of frames, are encoded in the same format, with the same aspect ratio, etc… It is utterly impractical in most cases and requires enormous CPU capacity to process, in many cases more than what is necessary for the actual transcoding / optimization.

Here is an example of a full reference video quality measurement method under evaluation and being submitted to ITU-T.

As a full reference approach, the model compares the input or high-quality reference video and the associated degraded video sequence under test. 

Score estimation is based on the following steps:

1)          First, the video sequences are preprocessed. In particular, noise is removed by filtering the frames and the frames are subsampled.

2)          A temporal frame alignment between reference and processed video sequence is performed.

3)          A spatial frame alignment between processed video frame and the corresponding reference video frame is performed.

4)          Local spatial quality features are computed: a local similarity and a local difference measure, inspired by visual perception.

5)          An analysis of the distribution of the local similarity and difference feature is performed.

6)          A global spatial degradation is measured using a Blockiness feature.

7)          A global temporal degradation is measured using a Jerkiness feature. The jerkiness measure is computed by evaluating local and global motion intensity and frame display time.

8)         The quality score is estimated based on a non-linear aggregation of the above features.

9)          To avoid misprediction in case of relatively large spatial misalignment between reference and processed video sequence, the above steps are computed for three different horizontal and vertical spatial alignments of the video sequence, and the maximum predicted score among all spatial positions is the final estimated quality score.

Reduced reference 

Reduced reference video measurement is performing the same evaluation as in the full reference model but only on a subset of the media. It is not widely used as frames need to be synchronized and recognized before evaluation.

No reference 

No reference video measurement is the most popular method in video optimization and is used usually when the encoding method is known. The method relies on the tracking of artefacts in the video, such as blockiness, jerkiness, blurring, ringing…etc. to derive a score.

Most vendors will create a MOS score from proprietary no reference video measurement derived from mathematical models. The good vendors constantly update the mathematical model with comparative subjective measurement to ensure that the objective MOS score sticks as much as possible to the subjective testing. You can find out who is performing which type of measurement and their method in my report, here.

Mobile video optimization 2012: executive summary

As I publish my first report (description here), have an exclusive glance with the below summary.

Executive Summary

V

ideo is a global phenomenon in mobile networks. In only 3 years, it has exploded, from a marginal position (less than 10%) to dominating mobile traffic in 2012 with over 50%.

Mobile networks until now, have been designed and deployed predominantly for transactional data. Messaging, email, browsing is fairly low impact and lightweight in term of payload and only necessitated speed compatible with UMTS. Video brings a new element to the equation. Users rarely complained if their text or email arrived late, in fact, they rarely noticed. Video provides an immediate feedback. Consumers demand quality and are increasingly assimilating the network’s quality to the video quality.

With the wide implementation of HSPA (+) and the first LTE deployments, together with availability of new attractive smartphones, tablets and ultra book, it has become clear that today’s networks and price structure are ill-prepared for this new era.

Handset and device vendors have gained much power in the balance and many consumers chose first a device before a provider.

In parallel, the suppliers of content and services are boldly pushing their consumer relationship to bypass traditional delivery media. These Over-The-Top (OTT) players extract more value from consumers than the access and network providers. This trend accelerates and threatens the fabric itself of the business model for delivery of mobile services.

This is the backdrop of the state of mobile video optimization in 2012. Mobile network operators find themselves in a situation where their core network is composed of many complex elements (GGSN, EPC, browsing gateways, proxies, DPI, PCRF…) that are extremely specialized but have been designed with transactional data in mind. The price plans devised to make sure the network is fully utilized are backfiring and many carriers are discontinuing all-you-can-eat data plans and subsidizing adoption of limited, capped, metered models. Radio access is a scarce resource, with many operators battling with their regulators to obtain more spectrum. The current model to purchase capacity, based on purchasing more base stations, densifying the network is finding its limits. Costs for network build up are even expected to exceed data revenues in the coming years.

On the technical front, many operators are hitting the Shannon’s law, the theoretical limit for spectrum efficiency. Diminishing returns are the rule rather than the exception as RAN become denser for the same available spectrum. Noise and interferences increase.

On the financial front, should an operator follow the demand, it would have to double its mobile data capacity on a yearly basis. The projected revenue increase for data services shows only a CAGR of 20% through 2015. How can operators keep running their business profitably? 

Operationally, doubling capacity every year seems impossible for most networks who look at 3 to 5 years roll out plans.

 Solutions exist and start to emerge. Upgrade to HSPA +, LTE, use femto cells or pico cells, change drastically the pricing structure of the video and social services, offload part of the traffic to wifi, implement adaptive bit rate, optimize the radio link, cache, use CDNs, imagine new business models with content providers, device manufacturers and operators… All these solutions and other are examined in this report.

Video optimization has emerged as one of the technologies deployed to solve some of the issues highlighted above. Deployed in over 80 networks globally, it is a market segment that has generated $102m in 2011 and is projected to generate over $260m in 2012. While it is not the unique solution to this issue, {Core Analysis} believe that most network operators will have to deploy video optimization as a weapon in the arsenal to combat the video invasion in their network. 2009 to 2011 saw the first video optimization commercial deployments, mostly as a defensive move, to shore up embattled networks. 2012 sees video optimization as a means to complement and implement monetization strategies, based on usage metering and control, quality of experience measurement and video class of service delivery.

GSMAOneAPI: One API to rule them all?

In June 2010, the GSMA released the specifications for its GSMAOneAPI. “A set of APIs that expose network capabilities over HTTP. OneAPI is developed in public and based on existing Web standards and principles. Any network operator or service provider is able to implement OneAPI.”

The API is based on xml/SOAP, its version 2, available since June 2011 includes SMS, MMS, Location and Payments as well as Voice Call Control, Data Connection Profile and Device Capability.

A live pilot implementation is ongoing in Canada with Bell, Rogers and Telus. It provides the capability for a content provider to enable cross network features such as messaging, call and data control. It is powered by Aepona.

The interesting fact about this API is that for the first time, it exposes some data control indication inherent to the core and RAN networks to potential external content providers or aggregators.

I went through an interesting demonstration on the GSMAOneAPI stand at Mobile World Congress 2012 by a small company called StreamOne, out of the Netherlands.

The company uses the API to retrieve from the operator the bearer the device is currently connected on. Additional extensions to the API currently under consideration by GSMA include download speed, upload speed and latency. These data points, when available to the content providers and aggregators could go a great way towards making techniques such as Adaptive Bit Rate more mobile friendly and potentially make way for a real bandwidth negotiation between network and provider. It might be the beginning of a practical approach to two sided business models to monetize quality of experience and service of OTT data traffic. As seen here, ABR is lacking capabilities to provide both operators and content providers with the control they need.

Of course, when looking at the standardization track, these efforts take years to translate into commercial deployments, but the seed is there and if network operators deploy it, if content providers use it, we could see a practical implementation in the next 3-5 years. Whant to know more about practical uses and ABR alternatives, check here.

NSN buoyant on its liquid net

I was with Rajeev Suri, CEO of NSN, together with about 150 of my esteemed colleagues from the press and analyst community on February 26 at Barcelona's world trade center drinking NSN's Kool Aid for 2012. As it turns out, the Liquid Net is not hard to swallow.

The first trend highlighted is about big data, big mobile data that is. NSN's prediction is that by 2020, consumers will use 1GB per day on mobile networks.
When confronted with these predictions, network operators have highlighted 5 challenges:

1. Improve network performances (32%)
2. Address decline in revenue (21%)
3. Monetize the mobile internet (21%)
4. Network evolution (20%)
5. Win in new competitive environment (20%)

Don't worry if the total is more than 100%, either it is was a multiple choice questionnaire or NSN's view is that operators are very preoccupied.

Conveniently, these challenges are met with 5 strategies (hopes) that NSN can help with:

1. Move to LTE
2. Intelligent networks and capacity
3. Tiered pricing
4. Individual experience
5. Operational efficiency

And this is what has been feeding the company in the last year, seeing sales double to 14B euros in 2011 and turning an actual operating profit of 225m euros. The CEO agrees that NSN is not back yet and more divestment and redundancies are planned (8,500 people out of 17,000 will leave) for the company to reach its long term target of 10% operating profit. NSN expects its LTE market share to double in 2012.

Liquid Net

Liquid networks is the moniker chosen by NSN to answer to the general anxiety surrounding data growth and revenue shrinkage. It promises 1000 times more capacity by 2020 (yes, 1000) and the very complex equation to explain the gain is as follow: 10x more cell sites (figures...), 10 times more spectrum and 10 times more efficiency.

The example chosen to illustrate Liquid net, was I think, telling. NSN has deployed its network at an operator in the UK where it famously replaced Ericsson last summer. It has been able since to detect devices and capabilities and adapt video resolutions with Netflix for instance that resulted in 50% less engorgement in some network conditions. That is hard to believe. Netflix being encrypted, I was scratching my head trying to understand how a lossless technique could reach these numbers.

The overall savings claimed for implementing liquid networks were 65% capacity increase, 30% coverage gain and 35% reduction in TCO.

Since video delivery in mobile networks is a bit of a fixation of mine, I decided to dig up more into these extraordinary claims. I have to confess my skepticism at the outset. I am familiar with NSN, having dealt with the company as a vendor for the last 15 years and am more familiar with its glacial pace of innovation in core networks.

I have to say, having gone through a private briefing, presentation and demonstration, I was surprised by the result. I am starting to change my perspective on NSN and so should you. To find out why and how, you will need to read the write up in my report.