基于语义空间统一表征的视频多模态内容分析技术

视频是数据处理中综合性能最高,包含内容最广的载体.视频题目通过文字表达,内容通过连续图像帧表达,另外部分视频还包含背景音乐或者解说旁白.因此,视频处理即是对文字、图像、声音的多模态处理.着眼于多模态处理技术,提出基于语义空间统一表征的视频多模态内容分析框架,利用多种架构的深度神经网络,对视频的文字、图像、音频进行分别处理,为达到统一的功效,将不同结构的深度神经网络归结到语义空间,通过语义空间进行综合认知.提出的架构清晰、层次分明,对于视频理解的建模具有指导意义.     

基于人工智能的跨媒体感知与分析技术研究

随着新媒体技术的飞速发展,传统方法已难以准确表达具有人工智能属性的复杂知识结构,跨媒体成为大家关注的焦点。当前,媒体数据感知与分析已经从文本、语音、图像以及视频等单一媒体形态向覆盖网络空间与物理空间的跨媒体融合转变。研究满足新一代人工智能发展规划的跨媒体感知和分析技术体系,并依托知识图谱、长短时记忆网络以及卷积神经网络等技术,实现多通道网络数据爬取、实体统一表征、文本语义识别以及视图像分类等,可有效支撑舆情分析、新闻追踪以及情报获取等领域的跨媒体应用。     

The future of online social networks (OSN): A measurement analysis using social media tools and application

The explosion of online social networks (OSN) has created an interactive and communicative global phenomenon that has enabled billions of users to connect to other individuals on Facebook and Twitter but also with media sharing platforms such as Instagram and Pinterest. This study examines the current use of social media platforms and explores the factors that help define the long term implications of social media. The study employed a nationwide survey collected from 2012 to 2013 and is available from the PEW Internet research center of more than 2000 American citizens’ behaviour towards OSNs. The results revealed strong predictors of OSN that form the connections among users; and the core significant predictors: age, gender and access to mobile Internet that foster the adoption and usage of OSN in the future. Furthermore, online activities such as posting video content on social networks also highlighted the online usage patterns and trends of using social media to actively engage with other users more willingly than text. This is due to the viral nature of online media sharing on social media and as part of the video viewing and creating experience. An outline of practical implications of the findings and areas for future research is also discussed.     

Managing QoS requirements for video streaming: from intra‐node to inter‐node

Streaming video over IP networks has become increasingly popular; however, compared to traditional data traffic, video streaming places different demands on quality of service (QoS) in a network, particularly in terms of delay, delay variation, and data loss. In response to the QoS demands of video applications, network techniques have been proposed to provide QoS within a network. Unfortunately, while efficient from a network perspective, most existing solutions have not provided end‐to‐end QoS that is satisfactory to users. In this paper, packet scheduling and end‐to‐end QoS distribution schemes are proposed to address this issue. The design and implementation of the two schemes are based on the active networking paradigm. In active networks, routers can perform user‐driven computation when forwarding packets, rather than just simple storing and forwarding packets, as in traditional networks. Both schemes thus take advantage of the capability of active networks enabling routers to adapt to the content of transmitted data and the QoS requirements of video users. In other words, packet scheduling at routers considers the correlation between video characteristics, available local resources and the resulting visual quality. The proposed QoS distribution scheme performs inter‐node adaptation, dynamically adjusting local loss constraints in response to network conditions in order to satisfy the end‐to‐end loss requirements. An active network‐based simulation shows that using QoS distribution and packet scheduling together increases the probability of meeting end‐to‐end QoS requirements of networked video. Copyright © 2005 John Wiley & Sons, Ltd.     

Virus infection control in online social networks based on probabilistic communities

With the rapid progress of the Internet, more and more people socialize and make new friends through online social network sites and applications, such as Facebook, Twitter, and MSN Messenger. The number of users of these online social network sites and applications has increased significantly within a short period. Unfortunately, online social networks are often the platforms that propagate malicious software, such as viruses. The malicious software is spread by infected users who automatically send fake requests to other users. After accepting the fake requests, the users are infected. In a realistic environment, everyone can decide whether to accept or decline a request, and thus, we study how to construct and use community structures to efficiently control virus propagation in online social networks whose users each have a probability, namely, inclination, of accepting a request. In this paper, a community detection algorithm is proposed to detect communities in online social networks. In addition, a number of users are selected from the communities for patching and controlling virus propagation. Simulation results show that our proposed method provides good performance in terms of the number of distributed patches.Copyright © 2013 John Wiley & Sons, Ltd.     

Mobile social networking: reconnect virtual community with physical space

Online social networks (OSNs) such as Facebook and MySpace, etc., greatly improve our social connectivity and collaboration. However, those applications lead to a shift from physical communities to virtual communities. The recent availability of mobile broadband connections and location technologies, their increasing affordability, and the usability of new mobile devices (e.g. smartphones) have led to the emergence of mobile social networks (MSNs), which re-connect the virtual community to the physical region, and move users between them in a way that enhances both. Currently, MSN applications are mushrooming and racing to replicate the success of social computing in the mobile domain. We argue that the potential success of MSNs lies in active collaboration among users, which naturally arises many interdisciplinary challenges. However, there exists no systematical survey about MSNs. This paper thoroughly characterizes the basic design principles, research architecture, typical techniques, and fundamental issues in MSNs from cross-discipline and application viewpoints. Our contributions lie in the following aspects: First, we summarized the basic design principles and fundamental issues that run through MSN researches and applications; then, from multidisciplinary viewpoint, the research architecture is divided into multi-dimensional structural characteristics and evolution of users’ rational behaviors. Finally, from application perspective, MSNs are categorized into two areas: Socially inspired mobile networking technologies, and enhanced real social life with mobile computing (people-centric tasks and place centric tasks). Briefly, this paper organizes the isolated topics and systems in existing work into meaningful categories, and structures the design space for identifying social-technical challenges, inspiring potentially interesting social networking applications, and suggesting important research opportunities.     

车联社会网络综述

车联社会网络是一种融合了社会科学与无线通信技术的移动通信网络,旨在为车载用户提供基于社会关系的数据投递服务。在分析国内外移动社会网络及车载网络研究现状基础上,提出了车联社会网络的基本架构及其典型应用,深入分析了车载网络与社会网络之间的关系,探讨了车联社会网络中信息投递、带宽分配、分布式存储和安全隐私等技术中存在的关键问题及其解决策略,最后讨论了该领域中新的研究方向与挑战。     

QoE‐based pseudo heartbeat message compression mechanism for future wireless network

Wireless multimedia services are increasingly becoming popular, which boosts the need for better quality‐of‐experience (QoE). However, there are many aspects leading to the degradation of real‐time video QoE, especially, a large number of always‐on‐line (AOL) applications existing in future wireless networks transmit heartbeat message periodically to keep always‐on, and hence induce heavy signaling costs and overload wireless networks. In this paper, we propose QoE‐based pseudo heartbeat message compression mechanism to reduce the number of signaling loads in the radio access network by intercepting the heartbeat message at a certain frequency in the proxy client. To maintain the protocol feature of the AOL applications, the heartbeat messages are reconstructed by the proxy server and sent to the application server. Furthermore, to analyze the influence of this mechanism on the video user, a new QoE perception model is proposed. Finally, combined the QoE perception model for video services with AOL services, the utility function for joint optimization multi‐services is developed to determine the optimum compression frequency. The simulation results show that the proposed mechanism greatly alleviates the signaling load and leads to a significant performance improvement on the QoE of video users, while a slight decrease in the QoE of AOL users. Copyright © 2016 John Wiley & Sons, Ltd.     

视频转码技术回顾

随着各种视频压缩标准的制定及计算机硬件、网络技术的发展，视频信息的应用呈现出爆炸式的增长。为了适应用户终端的多样性及网络自身的传输特性，20世纪90年代后期，视频转码成为视频编码领域一个新的研究热点。本文从视频转码的应用、体系结构和关键技术三个方面，回顾近几年来在视频转码方面的研究进展。     

基于模糊同质性映射的文本检测方法

视频图像中的文本是从语义层次对视频图像内容进行描述的非常有效信息,文本检测为基于语义的图像检索提供了条件。该文提出了一种基于模糊逻辑和同质映射相结合的文本检测方法,首先利用最大信息熵准则将原始图像模糊化;然后构造基于边缘信息和纹理信息的图像同质性,并利用它将图像映射到模糊同质性空间;最后在模糊同质性空间通过纹理分析检测文本区域。与直接在图像空间域中提取特征的文本检测方法相比,该方法对复杂背景视频图像的文本检测取得了更好的效果,并且适用于多种类型的视频图像中文本的检测。     

User identification via neural network based language models

Identifying compromised accounts on online social networks that are used for phishing attacks or sending spam messages is still one of the most challenging problems of cyber security. In this research, the authors explore an artificial neural network‐based language model to differentiate the writing styles of different users on short text messages. In doing so, the aim is to be able to identify compromised user accounts. The results obtained indicate that one can learn the language model on one dataset and can generalize it to different datasets to identify users with high accuracy and low false alarm rates without any modification to the language model.     

Streaming Multimedia for the Internet

Multimedia streaming enables Web sites to be enhanced with real-time audio and video content. Through the appropriate use of video compression technologies, audio and video content can be delivered directly into the Web browser environment to create a rich mix of text, graphics and multimedia. The audio quality and video window size available to viewers is primarily determined by network access speeds. For the Internet, audio and video can be successfully streamed at rates between 20 kbit/s and 37 kbit/s over a PSTN modem link. In the corporate intranet environment, where higher end-to-end network speeds exist, audio and video can easily be streamed at rates between 56 kbit/s and 450 kbit/s. This paper illustrates how commodity video compression techniques facilitate the introduction of real-time multimedia streaming over IP networks. It describes the technology components within a basic video streaming system and provides an overview of two applications that utilise this technology, namely Intranet TV and BT Results Webcast. The underlying network protocols required to support real-time Internet applications are also discussed and network measurement statistics obtained from Intranet TV and BT Results Webcast are also presented.     

A game‐theoretic model for users' participation in ephemeral social vehicular networks

Ephemeral social vehicular networks allow for short‐lived communications between occupants. While such transient interactions may provide important usage benefits such as traffic warnings, the reality of short‐lived interactions also poses challenges for deciding to participate in such social networks. In this paper, we develop a game‐theoretic model to understand the role of occupants' incentives to contribute to these ephemeral social networks. We rigorously study the properties of the model and resulting equilibrium outcomes. A further key objective of our work is to evaluate the robustness of our model to different notions of intrinsic motivation for network participation. As a baseline case, we assume that users are homogeneous regarding their motivation. Next, we investigate the impact of different motivational distributions and explore their impact on users' behaviors in a comprehensive simulation setup. We identify the circumstances under which ephemeral social networks are viable and can successfully sustain an increased level of network participation.     

The holy grail of content-based media analysis

Tools and systems for content-based access to multimedia and-image., video, audio, graphics, text, and any number of combinations-has increased in the last decade. We've seen a common theme of developing automatic analysis techniques for deriving metadata (data describing information in the content at both syntactic and semantic levels). Such metadata facilitates developing innovative tools and systems for multimedia information retrieval, summarization, delivery, and manipulation. Many interesting demonstrations of potential applications and services have emerged-finding images visually similar to a chosen picture (or sketch); summarizing videos with thumbnails of keyframes; finding video clips of a specific event, story, or person; and producing a two-minute skim of an hour-long program. In order to evaluate content-based research methodologies, the article considers intended users and whether alternative solutions exist and areas of research     

Interactive digital video networks: lessons from a commercialdeployment

Broadband networks are now being used in numerous trials throughout the world to provide interactive video services for residential customers. These trials and small-scale commercial deployments lack novel applications and services that make effective use of network capabilities and attract users to experiment with them. The rapid explosion of the Internet has also played a role in the current lack of interest and investment in broadband video networks. This article examines the lessons learned from the commercial deployment of a switched fiber-coax video dial tone network in Dover Township, New Jersey, USA. Future broadband networks will have to combine the capabilities of video dial tone networks and the Internet at an attractive price. Innovative end-to-end applications that go beyond cable TV and movies on demand are needed to renew commercial interest in residential broadband networks     

基于内容的视频语义分析关键技术

分析了视频数据与文本数据的差异,以及视频数据在视频分析检索方面存在的问题。从视频内容分析领域的研究热点出发,分别对视频语义库、与视频分析相关的视频低层特征、视频对象划分与识别、视频信息描述与编码等方面的技术进行了分析和对比。并提出了一个视频语义分析的框架和分析流程。     

On the characteristics of routing paths and the performance of routing protocols in vehicle‐formed mobile ad hoc networks on highways

Recently, intelligent transportation systems (ITS) is becoming an important research topic. One goal of ITS is to exchange information among vehicles in a timely and efficient manner. In the ITS research community, inter‐vehicle communications (IVC) is considered a way that may be able to achieve this goal. An information network built on the top of vehicles using IVC can be viewed as a type of mobile ad hoc networks (MANETs). In the past, several unicast routing protocols for MANET have been proposed. However, most of them are designed for general MANETs rather than for IVC networks. In this paper, we first used more realistic vehicle mobility traces generated by a microscopic traffic simulator (VISSIM) to understand the characteristics of routing paths in an IVC network. Based on the insights gained from the derived path characteristics, we designed and implemented an intelligent flooding‐based routing protocol for small‐scale IVC networks. Via several field trials conducted on highways, we compared the performance of ad hoc on‐demand distance vector (AODV) and our protocol. Our experimental results show that (1) our protocol outperforms AODV greatly in IVC networks and (2) our protocol can provide text, image, audio, and video services for small‐scale IVC networks (e.g., a platoon) quite well. Copyright © 2009 John Wiley & Sons, Ltd.     

Improving video QoE with IP over ICN

Information‐centric networking (ICN) has long been advocating for radical changes to the Internet, but the upgrade challenges that these entail have hindered its adoption. To break this loop, the POINT project proposed a hybrid, IP‐over‐ICN, architecture: IP networks are preserved at the edge, connected to each other over an ICN core. This exploits the key benefits of ICN, enabling individual network operators to improve the performance of their IP‐based services, without changing the rest of the Internet. This paper first provides an overview of POINT and outlines how it can improve upon IP in terms of performance and resilience. It then describes a trial of the POINT prototype in a production network, where real users operated actual IP‐based applications. As part of the trial, we carried out experiments to evaluate the quality of experience (QoE) for video services offered via either HLS or IPTV, using either IP or POINT as a substrate. The results from the trial verify that the IP‐over‐ICN approach of POINT offers enhanced QoE to the users of these video services, compared with traditional IP, especially under exceptional network conditions.     

Cache‐at‐relay: energy‐efficient content placement for next‐generation wireless relays

Uploading and downloading content have recently become one of the major reasons for the growth of Internet traffic volume. With the increasing popularity of social networking tools and their video upload/download applications, as well as the connectivity enhancements in wireless networks, it has become a second nature for mobile users to access on‐demand content on‐the‐go. Urban hot spots, usually implemented via wireless relays, answer the bandwidth need of those users. On the other hand, the same popular contents are usually acquired by a large number of users at different times, and fetching those from the initial content source each and every time makes inefficient use of network resources. In‐network caching provides a solution to this problem by bringing contents closer to the users. Although in‐network caching has been previously studied from latency and transport energy minimization perspectives, energy‐efficient schemes to prolong user equipment lifetime have not been considered. To address this problem, we propose the cache‐at‐relay (CAR) scheme, which utilizes wireless relays for in‐network caching of popular contents with content access and caching energy minimization objectives. CAR consists of three integer linear programming models, namely, select relay, place content, and place relay, which respectively solve content access energy minimization, joint minimization of content access and caching energy, and joint minimization of content access energy and relay deployment cost problems. We have shown that place relay significantly minimizes the content access energy consumption of user equipments, while place content provides a compromise between the content access and the caching energy budgets of the network. Copyright © 2015 John Wiley & Sons, Ltd.