Research Areas

Wireless Communications

The research fields mainly focus on Cognitive Radio technology, Polarization enable Spectrum Sharing, the wireless transmission and information processing under Cognitive Vehicular Networks.

In terms of cognitive radio technology, the research concentrates on spectrum sensing, allocation, access and power control, etc, the application and development of cognitive radio system under 5G/5G+ or the future networks, the experimental verification platform for spectrum sharing in cognitive radio network, which is based on USPR and GNU radio.

As for the research in Polarization enabled Spectrum Sharing, in order to discover the potential spectrum opportunity, the research mainly focuses on the analysis of wireless channels’ depolarization and modeling theory, instantaneous identification and analysis of polarization status, sensing methods of the primary users’ polarization status, changed polarization of receiving signal and transmitting signal, etc.

Among the appealing field of Cognitive Vehicular Networks, which combines the cognitive radio technology with vehicular communication, the main research comprises data collection, procession and dissemination, data mining technology in big data domain which can process the massive data collected in vehicular networks with high efficiency, and some other key research points, like controlling channel access, spectrum sensing and performance analysis, spectrum sharing under the vehicle networking scenario.

Visible Light Communication

  • VLC prototype based on IEEE 802.15.7
  • We establish a USRP basedprototypethat is compatible with IEEE 802.15.7. At the transmitter, a PC is used to generate the source data, which is followedby the USRP. In USRP, the signal processing is accomplished and the signal is outputted to the LED driver. The LED driver amplifies the useful signal and drives LED transmitting the useful signal. At the receiver, a photodiode is employed to transform the optical signal into voltage, which is then capturedby an oscilloscope and is conveyed to the receiver USRP for demodulation as well. Finally, the demodulated signal is outputted to a PC. This platform can transmit high definition video in real time and data rate as high as 2.5 Mbps can be achieved.

  • VLC platform for vehicular applications
  • A camera based VLC platform intended for vehicular application is implemented on a 1:25 scale model.There are three vehicles in the implemented prototype, where Vehicle 1 and Vehicle 2 are transmitters.As the green arrows shows, Vehicle 1 moves along the vertical direction while Vehicle 2 moves along the horizontal direction. Vehicle 3 equipping a camera acts as receiver.The case that vehicles passing through a crossroad is simulated by Vehicle 1 and Vehicle 3 while the case that vehicles moving with the same direction is simulated by Vehicle 2 and Vehicle 3. In this platform, specialized LED tracking and environment sensing blocks are designed, which can achieve reliable VLC under strong interference noise and the rapid movement of vehicle.

  • The novel spatialdimming scheme
  • We propose a new type of dimming control scheme for visible light communications, which is termed as spatial dimming. Spatial dimming is a competitive scheme of conventional analgoue and digital dimming schemes. The basic idea of spatial dimming is controlling the activation of LEDs to achieve the required illuminaiton level. Since this will not change the form of transmitted signal, the distortion caused by spatial dimming is much lower than that of its counterparts. Therefore, it is able to obtian better communication and illuminaiton performance. In addition, it can be easily combined with other spatial domain based techiniques to further enhance its performance. Furthermore, we have proved that spatial dimming scheme will not compromise the uniform illuminaiton. We have also investigated its combination with multi-carrier modulation scheme and multi-user MIMO.

  • Indoor Location with multi-directional PDs and LEDs
  • To overcome the challenges that at least three LEDs are required for location, algorithms based on multi-directional PDs and multi-directional are proposed for VLC. The greatest strength of the proposed schemes is that only one LED or PD is required to achieve location service with centimeter accuracy.

Social Multimedia Computing

  • Representation Learning for Heterogeneous Information Network
  • Most real-world complex systems take the form of heterogeneous information networks where nodes/edges are of different types, such as social networks, event data, communication networks and knowledge graphs. This research focuses on representation learning for heterogeneous information network with the goal of mapping different types of nodes (such as text, image and video)to a common latent space, in order to support subsequent network processing and analysis tasks such as node classification, link prediction and node visualization.

    Social-guided Representation Learning for Images via Deep Heterogeneous Hypergraph Embedding (ICME 2018, Best paper diamond award)

    Event2vec: Heterogeneous Hypergraph Embedding for Event Data

    (ICDMW 2018)

  • Social-sensed Cross-media Computing
  • With the social trend in multimedia data generation and consumption, media data which crosses different information sources, diverse modalities, and various characteristics of properties, such astext, image, video and GPS information are generated every day and easily obtained in social networks. We study theories and methods in social-sensed cross-media computing with the aim of effectively mining valuable knowledge from huge and heterogeneous data sets. We also design a social-sensed cross-media visualization and application platform, which can conduct opinion analysis, influence analysis and prediction, eventsummarization, event reconstruction, etc.