Search

Person re-identification via 3D skeletons is an emerging topic with great potential in security-critical applications. Existing methods typically learn body and motion features from the body-joint trajectory, whereas they lack a systematic way to model body structure and underlying relations of body components beyond the scale of body joints. In this paper, we for the first time propose a Self-supervised Multi-scale Skeleton Graph Encoding (SM-SGE) framework that comprehensively models human body, component relations, and skeleton dynamics from unlabeled skeleton graphs of various scales to learn an effective skeleton representation for person Re-ID. Specifically, we first devise multi-scale skeleton graphs with coarse-to-fine human body partitions, which enables us to model body structure and skeleton dynamics at multiple levels. Second, to mine inherent correlations between body components in skeletal motion, we propose a multi-scale graph relation network to learn structural relations between adjacent body-component nodes and collaborative relations among nodes of different scales, so as to capture more discriminative skeleton graph features. Last, we propose a novel multi-scale skeleton reconstruction mechanism to enable our framework to encode skeleton dynamics and high-level semantics from unlabeled skeleton graphs, which encourages learning a discriminative skeleton representation for person Re-ID. Extensive experiments show that SM-SGE outperforms most state-of-the-art skeleton-based methods. We further demonstrate its effectiveness on 3D skeleton data estimated from large-scale RGB videos. Our codes are open at https://github.com/Kali-Hac/SM-SGE., Accepted at ACMMM 2021 Main Track. Sole copyright holder is ACMMM. Codes are available at https://github.com/Kali-Hac/SM-SGE

People spend up to one-third of lives asleep, and healthy sleep habits can make a big difference in their quality of life. But in modern society, many people have unhealthy sleep diaries and suffer from various sleep disorders, which may result in irregular mood fluctuations or even mental health problems such as anxiety and depression. We propose the Emotion-Aware Smart Tips (EAST), a novel approach that could help to inform users about their irregular emotional states with smart tips to improve their sleep qualities. EAST aims at helping users keep healthy sleep schedules and emotional states by providing smart tips through a novel model that combines multivariate regression, random forest, and neural network to quantify the relations between sleep patterns and emotional states. Prototype implementation and initial experiments of EAST in mobile phones have demonstrated its desired functionality and practicality for real-world deployment.

Road safety is a huge concern due to the large number of fatalities and injuries caused by road accidents. Research has shown that fatigue can adversely affect driving performance and increase risk of road accidents. It has been shown that driving performance is enhanced by stress-relieving music which thereby promotes safer driving. Context-aware music delivery systems promote safer driving through intelligent music recommendations based on contextual knowledge. Two key aspects of situation-aware music delivery are effectiveness and efficiency of music recommendation. Efficiency is a critical aspect in real-time context based music recommendation as the music delivery system should quickly sense any change in the situation and deliver suitable music before the sensed context-data becomes obsolete. We focus on the efficiency of situation-aware music delivery systems in this paper. Music mood-mapping is a process which helps in understanding the mood of a song and is hence used in situation-aware music recommendation systems. This process requires a large processing time due to the complex calculations and large sizes of music files involved. Hence, optimizing this process is the key to improving the efficiency of context-aware music delivery systems. Here, we propose a novel cloud and crowd-sensing based approach to considerably optimize the efficiency of situation-aware music delivery systems.

The heterogeneous cloud radio access network (H-CRAN) has been emerging as a cost-effective solution supporting huge volumes of mobile traffic in the big data era. This paper investigates potential security challenges on H-CRAN and analyzes their likelihoods and difficulty levels. Typically, the security threats in H-CRAN can be categorized into three groups, i.e., security threats towards remote radio heads (RRHs), those towards the radio cloud infrastructure and towards backhaul networks. To overcome challenges made by the security threats, we propose a hierarchical security framework called Secure H-CRAN (SH-CRAN) to protect the H-CRAN system against the potential threats. Specifically, the architecture of SH-CRAN contains three logically independent secure domains (SDs), which are the SDs of radio cloud infrastructure, RRHs and backhauls. The notable merits of SH-CRAN include two aspects: (i) the proposed framework is able to provide security assurance for the evolving H-CRAN system, and (ii) the impacts of any failure are limited in one specific component of H-CRAN. The proposed SH-CRAN can be regarded as the basis of the future security mechanisms of mobile bag data computing.

Nowadays more and more organizations focus on reducing traffic accidents and defensive measures for safe driving. The vigilance level (e.g., negative emotion and fatigue) also accounts for the road injuries. Till now, there is no systematic solution for different mobile devices that can effectively infer the mood and fatigue of drivers in real-time or conveniently be used by drivers, nor incentive scheme for drivers in large scale to stimulate their positive and secure driving collaboratively with friends in a social context. In this paper, we propose the Mood-Fatigue Analyzer (MFA), a systematic solution that can be used in different middlewares on mobile devices, which can transform the data from sensors to context-aware mobile sensing applications for safe driving. The MFA employs multidimensional methods to get the drivers' real-time mood and fatigue information by sensors using the Internet of Things (IoT) deployed in and out of cars. Besides promoting safe driving with integrated sensors, the MFA could be built on a multi-tier vehicular social network (VSN) platform, which enables communication among drivers in a social context via cloud platform. Architecture implementation and experimental results of the MFA have demonstrated its desired functionalities and efficiency in drivers' daily lives and real-world deployment.

Advances in sensor cloud computing to support vehicular applications are becoming more important as the need to better utilize computation and communication resources and make them energy efficient. In this paper, we propose a novel approach to minimize energy consumption of processing a vehicular application within mobile wireless sensor networks (MWSN) while satisfying a certain completion time requirement. Specifically, the application can be optimally partitioned, offloaded and executed with helps of peer sensor devices, e.g., a smart phone, thus the proposed solution can be treated as a joint optimization of computing and networking resources. Our theoretical analysis is supplemented by simulation results to show the significance of energy saving by 63% compared to the traditional cloud computing methods. Moreover, a prototype cloud system has been developing to validate the efficiency of sensor cloud strategies in dealing with diverse vehicular applications.

Driving is an integral part of our everyday lives, but it is also a time when people are uniquely vulnerable. Poor road condition, traffic congestion and long driving time may bring negative emotion to drivers and increase the chance of traffic accidents. We propose SAfeDJ, a situation-aware in-car music delivery application, which turns people's trips into pleasant journeys and driving into a safe and enjoyable activity. SAfeDJ aims at helping drivers to diminish fatigue and negative emotion. It is built on a vehicular healthcare platform that enables communications among drivers and integrates with multiple types of sensors to promote safe driving. Prototype implementation and initial results of SAfeDJ have demonstrated its desired functionality in drivers' daily lives and feasibility for real-world deployment.

This paper presents TripleS, a novel mobile crowdsensing system enhanced by social networking services, which enables mobile users to participate and perform mobile crowdsensing tasks in an efficient manner. TripleS provides a flexible and universal architecture across mobile devices and cloud computing platforms by integrating the service-oriented architecture with multi-agent frameworks for mobile crowdsensing, with extensive supports to application developers and end users. The customized platform of TripleS enables dynamic deployments and collaborations of services and tasks during run-time of mobile devices. Our practical experiments show that TripleS performs its tasks with a considerable computation efficiency, and low computation and communication overhead on mobile devices. Also, the mobile crowdsensing application developed on TripleS demonstrates the functionalities and practical usage of TripleS.

Transport of information in vehicular clouds faces challenges due to intermittent connectivity and the fact that the already existing Internet protocol based transport solutions do not exploit the semantics of information to utilize the available contextual information. With the advent of Internet of Things and Machine to Machine communications, availability of contextual information through the wisdom of the crowd and ubiquity of sensors and devices calls for a shift towards networking of information beyond Internet Protocol (IP) level connectivity. We propose a novel approach for forwarding and discarding policy that can be utilized by content aware network elements. The proposed method makes use of multidimensional scaling techniques that leverages the spectral characteristics of information predicates. By evaluations and analysis we show that by considering the networking of information paradigm for vehicular clouds our proposed clustering technique yields a lower processing cost and complexity as the system scales.

This paper presents Social Drive, a novel crowdsourcing-based vehicular social networking (VSN) system for green transportation. Social Drive integrates the standard vehicular On-Board Diagnostics (OBD) module, leverages the advantages of cloud computing and popular social networks, and incorporates a novel rating mechanism about the fuel economy of drivers. Based on these, Social Drive provides a user-friendly mobile application on smartphones targeting drivers, which enables a seamless and economic solution that promote drivers' awareness of their driving behaviors regarding fuel economy of specific trips. Our practical experiments have demonstrated that Social Drive works efficiently with low battery consumption and low networking overhead on popular mobile devices.

This paper proposes a novel service-oriented vehicular social networking platform called VSSA for transportation efficiency. VSSA consists of two layers: service layer and application layer. The service layer is implemented based on the service-oriented architecture (SOA) to ease application developments. By extending and composing the web services in this layer, application developers can efficiently and flexibly develop different new functions or applications for transportation scenarios. The application layer is oriented to vehicular users and it integrates with functions of mobile social networks. With dynamic and automatic service collaboration support, it enables people to easily collaborate and help each other through their mobile devices in transportation situations. Moreover, VSSA provides a context-awareness mechanism to support automatically and intelligently predicting the potential incoming traffic congestions.

This paper proposes a semantic-based multi-agent framework to support development of vehicular social network applications. In the programming model of the framework, software platforms of vehicular social network systems can be developed by the collaboration of mobile agents and service (or resident) agents, where resident agents provide application services on devices and mobile agents provide communication services on behalf of owner applications. On top of the device infrastructure, the architecture of the proposed framework consists of three layers: framework service layer, software agent layer and application layer, to fully support dynamic and collaborative tasks of vehicular social networks. The multi-layer architecture design of the framework fully supports self-adaptive applications in vehicular social network environments, and is readily extensible to support new features. Developers can easily and effectively develop diverse applications for the vehicular social networks.