The need for mobile connectivity has grown dramatically, spurring developers to bring implementations to market as quickly as possible. Nowhere is this more evident than among service providers and enterprises.

Despite this advancement in mobile communications technology, service providers and enterprises are not consistently adhering to standards, and users are often forced to endure video conferencing experiences that leave much to be desired in terms of audio and visual quality. The reason for this is obvious. Since the introduction of the World Wide Web, the call to deliver ever-larger data packets via e-mail or Internet downloads has prompted companies to roll out networks capable of carrying these increased loads.

Video services, however, have remained on the periphery until recently. Today's market realizes the business value in using video conferencing, media servers, video ring tones, video mail, and similar services in the office as well as on the road. These services enable staff, suppliers, and customers to operate within a seamless, unified communications environment.

As the market has embraced video, many developers are now creating both video and IP telephony applications for use over high-bandwidth networks.

Test VQ readiness first, then deploy

These applications do not always perform as expected because, as mentioned previously, most corporate networks are built and optimized to carry data, not video.

Besides affecting enterprises, poor user experiences are hampering carriers who have invested billions of dollars to develop 3G-capable networks and handsets and yet failed to generate significant user uptake.

Both enterprises and carriers have overlooked the need to determine a network's VQ readiness when developing video and launching new video services and VoIP applications. By incorporating software tools that enable VQ testing, companies can minimize development work and time to market as well as free developers to deliver products that truly leverage the power and promise of video conferencing.

Determining VQ

VQ is a significant measure of a network's reliability, requiring a comprehensive VQ calculation to be included with any profound network investigation. However, given that a detailed network check is an extensive undertaking that involves detailed planning and in-depth testing on all parameters, VQ calculation becomes unfeasible without using a third-party software tool. Service providers and enterprises must take this into account when seeking to obtain reliable and objective results about a network's ability to deliver top-quality video conferencing between staff and third parties, regardless of their location or endpoint devices.

Video-related network parameters such as bit rate, frame rate, packet loss rate, jitter, and delay can affect VQ. Low network performance reduces the Mean Opinion Score (MOS, a measure of perceived video quality) by introducing video artifacts such as blurring and blockiness.

Artifact types and their severity depend on video coding parameters such as codec type and resolution. For example, low bit rate occasionally leads to blockiness in video coded with H.263. Another example is packet loss, which sometimes results in local distortion within a high-quality frame.

VQ can be designed to consider a Full Reference (FR), Partial Reference (PR), and No Reference (NR) to the distorted video sequence. The most generic mode is NR, which enables VQ for endpoints that are incapable of storing video references such as 3G cellular phones. Moreover, when VQ is intended for network checkup, it is best to assume the NR mode because even if the endpoint can produce a reference, mid-network components such as MCUs might modify video synchronization and scaling, therefore disabling the functions for matching and comparing video to its reference.

However, at the end of the day, it is the user who experiences the product. Research shows that VQ software reliability can only be accurate when correlated with an MOS obtained from tests conducted on users in the field. This might seem obvious; after all, many of the products consumers use every day have undergone exhaustive testing from both technical and user points of view. Consequently, VQ software reliability is determined by its correlation to human perception.

Testing for perception

To ensure optimal network performance, RADVISION offers a VQ-MOS program that combines results from its ProLab software add-on and its Human Vision VQ survey (an online and offline NR-based strategy).

Designed around internationally accepted test methods used by private industries, public agencies, and standards bodies, this survey requires subjects to view video sequences and grade them according to a single stimulus quality scale method. Each review runs for 30 minutes, preceded by a short sample session demonstrating high- and low-quality videos. Five sets of homogenous video sequences with gradual decreases in video quality are used to familiarize subjects with testing procedures. Video sequences must be homogenous to prevent MOS bias due to content, and numerous subjects with different skills and expertise are required to achieve optimal results.

VQ results are obtained using unique video content specifically designed for video-telephony network checking. Video sequences for codec types H.264, MPEG-4, and H.263 (using several different implementations for each codec type); resolutions Common Intermediate Format (CIF), Quarter CIF (QCIF), and HD; bit rates 56-2,048K; frame rates 8-30; and packet loss rates 0-10 percent are used to produce the best results. Figure 1 illustrates a decrease in VQ scores resulting from an increase in network packet losses for an H.264 video at CIF resolution. Note that the graph correlates with well-known trends in results from tests using human subjects.

Optimized video performance

VQ testing tools coupled with end-user feedback from a testing scenario such as the Human Vision VQ survey can help enterprises improve the quality of video conferencing and IP telephony systems as well as increase user uptake.

By combining results from objective and subjective testing devices, companies can gain the information needed to optimize network performance for video deployments. This integrated approach enables developers to isolate operability issues in real time and monitor network quality, thus allowing them to identify problems earlier in the development process, eliminate silos and blind spots, and ensure high-quality user experiences.

Eli Cohen is product manager for RADVISION (Tel Aviv, Israel) and founder of the ProLab Testing Suite. For the past nine years, he has created and managed innovative testing systems for VoIP and 3G, including VQ measurement. Before joining RADVISION in 1999, Eli maintained various project management positions in Israel and France for more than 11 years. He holds a BSc in Computer Science from Jerusalem College of Technology.

RADVISION, Ltd.
+972-3-767-9300
[email protected]
www.radvision.com