Interview with a faculty candidate for the Computer Engineering Degree

Time: 11-12 --  All are welcome!

Title: Multimedia Personalization and Delivery for Mobile Clients in Resource-Constrained Environments

Presenter: Yong Wei, Department of Computer Science, The University of Georgia

Abstract

The current proliferation of mobile computing devices and network 
technologies has created enormous opportunities for mobile device users to
communicate with multimedia servers and with one another. One of the 
natural limitations of handheld computing devices is that they are 
constrained by their resources, such as battery power capacity, viewing 
time limit, amount of data received, and in many situations, by available 
network bandwidth connecting these devices with multimedia servers. Thus, 
the original videos often need to be personalized or adapted in order to 
fulfill the clientUs request while simultaneously satisfying various 
client-side system-level constraints. We develop a mobile client-centered 
multimedia adaptation and delivery system which can optimally fulfill the 
client requests while simultaneously ensuring optimal utilization of the 
server-side and client-side system-level resources.

The client-centered multimedia adaptation and delivery system consists of 
the following five subsystems: (1) the video preprocessing subsystem, (2) 
the multiple-level video summarization and hierarchical content 
representation subsystem, (3) the video personalization subsystem, (4) the
multiple client request aggregation subsystem and (5) the client-side 
energy-aware multimedia streaming subsystem.

The video preprocessing subsystem performs temporal video segmentation and
indexing in the temporal domain. A data-driven stochastic modeling approach
is proposed to perform both video segmentation and video indexing in a 
single pass automatically.  Indexed video segments are then fed to the 
multiple level video summarization and hierarchical content representation
subsystem to build a visual content database. Each indexed video segment is
summarized at multiple levels of abstraction. Content-aware key frame 
selection algorithm and motion panorama computation are used to generate 
video summaries for each indexed video segment. In the video 
personalization subsystem, the client video content preference is matched 
with the video summaries stored in the visual content database. In order to
generate the personalized video summary, the client usage environment and 
the client-side system-level constraints are evaluated. The personalization
engine selects the optimal subset of video contents that are most relevant
to the clientsU preference(s) subject to the constraints imposed by the 
client. In order to utilize the server-end resources efficiently, a 
multiple-stage client request aggregation strategy is designed and 
implemented to aggregate similar client requests together such that the 
number of requests the multimedia server needs to process is reduced. The 
client-side statistical prediction-based multimedia streaming strategy 
reduces the wireless network interface card (WNIC) energy consumption to 
receive multimedia streams by judiciously transitioning the WNIC to a lower
power consuming sleep state during the no-data intervals in the multimedia
stream, without explicit support from the multimedia servers themselves.