Cisco
The contact email for John Bain is jbain@cisco.com
Finalised brief
(old) discussions
Introduction
In 2011, Cisco Systems offered an undergraduate group project to design and implementation an on-line testing system, involving the design and implementation of a new programming language.
The six students that worked on the project produced a complete and effective system which is now being considered for live use by another company. Hoping to continue the success of last year, Cisco Systems – represented by John Bain and Edd Inglis – are happy to put forward the following idea for another project.
Background
The Cisco site in Langley specialises in multipoint video-conferencing systems, designing hardware and associated software to allow many video endpoints to communicate in large-scale conferences. One key consideration in the design of such systems is to minimise the latency added to the video and audio, since this can have a marked effect on the naturalness of communication.
Many software solutions exist for measuring various aspects of the system latency, but there is no substitute for end-to-end (often called glass-to-glass, referring to the lens of the camera and display screen) measurements of video and audio latency. This type of measurement has traditionally been subjectively measured by humans, and is hence prone to error and bias.
Proposal
We would like a team to create a system for measuring end-to-end latency of video and audio in a point-to-point or multi-point video conferencing call. We envisage such a system will include a light emitter and matching light sensor, and a sound generator and receiver, though any equivalent ingenious solution would be equally applicable.
The system would obviously need to be operable on systems closely located in a testing environment, but consideration should be given to testing between geographically separated devices. This is illustrated by the diagram below, where perhaps endpoint A is in London and endpoint B in San Jose.
Equipment
We imagine this system could be developed in a number of different ways. Our anticipation is that a small, embedded device would provide the most appropriate starting point for this, but are happy to entertain other suggestions. Assuming an embedded platform is chosen, we are happy to supply the appropriate hardware - for example an Arduino clone or similar development board - or for students to use any other platform they may be familiar with from other parts of their course.
It should be possible to do complete testing of such a system using PC-based video endpoints, but we can also supply ‘real’ endpoints and conferencing infrastructure for testing towards the latter stages of the project if desired. (We anticipate an embedded platform will need the following peripherals at a minimum: ADC/DAC x2, GPIO, Network (Ethernet). ]
Notes to supervisors
We feel this project provides a number of different areas of investigation and interest. Some of these include:
- Real-time operation; by its nature, this is a time-critical problem. A successful system could either be implemented at a logic level, by using an off-the-shelf RTOS, or even with an RTOS developed by the students
- Hardware and software; our engineers work with both day-to-day, meaning the project is realistic as well as more widely scoped. The system will solve a genuine issue, potentially offering more interest than a toy problem.
- Network enabled; interacting with IP networks is challenging and interesting, and will allow students to use network analysis tools to examine existing or new protocols for communication between remote devices.
- Time synchronisation; there is a need to synchronise remote devices. This could involve the use of NTP or other measurements of the round-trip time of the network, to allow accurate measurement of the true glass-to-glass latency.
- Robustness; the system needs to be robust in the face of noisy and/or corrupted audio and video, as well as lossy networks between remote measuring devices. This offers the opportunity for students to investigate methods of making the system robust to errors and external network factors.
