|Title:||Live video identification and transmission over wireless network|
Wireless communication systems.
Real-time data processing.
Hong Kong Polytechnic University -- Dissertations
|Department:||Department of Computing|
|Pages:||xvi, 119 p. : ill. ; 30 cm.|
|Abstract:||With the advancement in mobile computing technology, the capture devices and storage for video content become more and more mature and convenient. Live video identification and transmission are two prevalent and fashionable topics in mobile multimedia computing area. The growth of online video content raises new opportunities for the processing and delivery of the contents. Compared with traditional video applications, unprecedented challenges are raised on mobile real time video identification and transmission. Massive data are created on network every day, especially for the video uploading and downloading from mobile devices to the cloud. Moreover, limited resources in the mobile wireless network, such as bandwidth and computation capacity, demand more efficient and effective approaches of real time video identification and transmission. In the interests of achieving multimedia identification and transmission higher accuracy in real time under limited resources, we proposed corresponding solutions to solve the two fundamental problems. Simultaneously, based on the proposed methodology, a mobile based multimedia computing system is needed so as to embed complex multimedia computing process and realize it in real time. This thesis consists of three parts. The first part focuses on video transmission scheduling, which is to schedule and distributed the multimedia resources to all the users according to the multimedia unique characters. The second part focuses on video-based human action recognition, which is to exact the specific feature of the action and construct a model for action recognition. In the third part, a Mobile Cloud Computing (MCC) system is developed and supported by the above mentioned technologies, namely Real-time mobile based Video Recognition (RVR) system. The system demonstrates the live video transmission and identification can be processed in real time over wireless network with high accuracy under scarce resources. Firstly, we study the problem of video transmission scheduling. We propose a newly video delivery scheme, Utility Coordination Function (UCF), over 802.11 networks. Given the limited wireless resources, supporting multi-user video streaming with good video Quality-of-Service (QoS) is very challenging. The key difficulties involve providing good playback quality while also satisfying the stringent video packet delay bounds, especially for transmitting large amount of data under the limited resources such as bandwidth. The allocation of wireless resources needs to be efficient and coordination of mobile video users should have a distributed fashion. In this thesis we present a distributed framework for multi-user video streaming over an ad-hoc 802.11 like wireless networks. The proposed algorithm is based on a utility-driven mechanism that adjusts the video users' sending rates according to Application layer video buffer status. We propose multiple schemes towards different levels of user requests, and deal the problem with scalable techniques. Simulation results demonstrate that the proposed scheme is quite efficient on radio resource will have better QoS than content blind Distributed Coordination Function (DCF) scheme. Besides, we also prove that our proposed solution is robust against the possible variations in the network.|
Secondly, we study the problem of video-based human action recognition. We propose a spline approximation approach for video based action recognition to deal with large scale database. Video action recognition is another active research topic in computer vision and communication. Effective and fast processing approaches are highly demanded. Traditional pattern recognition and machine learning techniques can solve problems for text and image with satisfactory performance. However they become less helpful when processing large amount of video data. Besides, some statistic models designed for some special video processing applications, cannot handle the general video-based pattern recognition problem. In this thesis, we have tackled these problems from several aspects including simplifying the video representation and dimensionality reduction, improving spatio-temporal modeling, and speed up the online matching issue. The proposed approach focused on merging the current training trajectories into a much smaller but discriminative dataset to accelerate the processing for matching. An extension is also considered by introducing the idea of graph embedding; we polish the subspace learning with constructing an affinity matrix, to better evaluate the similarity within the same class during the training session. Experimental results demonstrate the proposed methods work effectively and efficiently. Thirdly, we proposed a Real-time mobile based Video Recognition (RVR) System over Mobile Cloud Computing (MCC). The cloud computing and mobile computing technologies lead to the newly emerging MCC paradigm. Three major approaches have been proposed for mobile cloud applications: 1) extending the access to cloud services to mobile devices; 2) enabling mobile devices to work collaboratively as cloud resource providers; 3) augmenting the execution of mobile applications on portable devices using cloud resources. This part focuses on the third approach in supporting mobile data stream applications by employing the proposed transmission and recognition algorithms. More specifically, we apply the optimized partitioning algorithms to the RVR system, which separates the computation of a real time video application between the cloud and mobile devices and then achieves maximum speed in processing the streaming data under predefined recognition accuracy. We first involve a real time partition algorithm for MCC based live video recognition system. Both numerical evaluation and real world experiment have been performed, and the results show that the proposed system can achieve better performance in terms of throughput than without employing the proposed algorithms.
|Rights:||All rights reserved|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: