We are devoted to undertaking impactful disciplinary research, keeping up with the knowledge advancement and global development in computing and information technology over the past four decades.
Our strengths:
Human eyes are not only the organ to observe the world, but also the window into health. The rich information about retinal vessels offers the vital details that would enable doctors to inspect the health conditions and examine retinal disorders caused by some systemic diseases such as diabetes, hypertension and arteriosclerosis. According to the World Health Organization, the number of diabetic patients is increasing rapidly and diabetes has remained a major cause of disability and death because of its various complications. It is crucial to detect diabetes and its complications at early stage for proper and timely treatment. Diabetic retinopathy (DR) is one of the common complications of diabetes, which causes damage to the retina of the eye and could eventually lead to blindness. However, DR often has no early warning signs. Conventional ophthalmology-based approach to DR detection imposes heavy workload on doctors and does not provide information about the progress of diabetic complications and the effect of therapy.
It has become international efforts to detect various diabetic complications timely, efficiently, safely, conveniently at low cost. To achieve the objectives of effective diabetic care with high detection accuracy rate and short processing time, the research team led by Prof. Jane You has developed a new retinal imaging system for computer-aided non-intrusive detection of diabetic retinopathy with a systematic mechanism. The new system allowed specialists to study lesions by providing a multi-level detection platform to handle large collection of image data easily without human error, speed up the process for mass screening, facilitate information sharing among medical experts and automate data analysis with high accuracy rate. In addition, the system is also coupled with information security and privacy protection for high-performance. The prototype system which can identify a retinal image with DR above 90% accuracy received positive feedback for practical use from doctors during its clinical trials at different hospitals. Currently the team is collaborating with doctors to develop an e-EyeGuard system in mobile cloud computing environment as an application to telemedicine and e-healthcare.
The research results of this project at different stages led to three consecutive general research funds (GRF) from the Hong Kong government during 2007-2009 and one innovation technology fund (ITF) in 2009. The deliverables out of these funded projects include research publications in top journals such as IEEE Trans. Medical Imaging and Pattern Recognition, a prototype system which was licensed to Wealth Billion International (HK) Ltd in 2011, two Ph.D graduates in 2011 and 2013 respectively. The research work also gained international recognition which is evidenced by the 2nd place in international competition (SPIE Medical Imaging’2009 Retinopathy Online Challenge ROC’2009), the special prize and gold medal with jury’s commendation at the 39th International Exhibition of Inventions of Geneva in 2011 and ICMLC2011 Lofti Zadeh best paper award (The 2011 International Conference on Machine Learning and Cybernetics). The research findings are currently pending on a US patent.
Currently, parallel database is the most high-end data analytical system in the big data analytical system market. Examples include HP’s Vertica (used by Obama’s election team during 2012 US President Election) and SAP’s HANA (used by Germany soccer team during 2014 World Cup). Rather than installing and using parallel database locally in a company, Parallel-Database-as-a-Service (PDaaS) can be provided to move much of the operational burden from the company to their cloud servers. One of the key challenging issues, however, is how to operate a PDaaS using less resource (e.g., number of computers, electricity) so as to minimize the operation cost. Solving the problem would greatly promote the acceptance of big data analytic tools in the market. Consequently, the industries have a stronger incentive to investigate the benefits of using big data in their business.
The research team has been dedicated to developing advanced techniques to reduce the operation cost of a PDaaS. Before their invention, the traditional approach is to let each PDaaS client obtain an exclusive amount of resources from the cloud servers, which is not cost-effective because the resources are wasted when some clients are idle. The team has successfully developed techniques to let all PDaaS clients share the common pool of computing resources in a PDaaS, while avoiding resource contentions when many clients are active. With this innovative technology, PDaaS providers can significantly reduce the resource requirements of operating a PDaaS by up to 80%. This enables PDaaS providers to significantly reduce the service price, which could be a strong catalyst to push the various industries to use big data in their business.
The conditions of civil structures like skyscrapers and long bridges are critical for public safety. Existing systems for monitoring the health condition of civil structures are wired and need to lay cables and use centralized control, incurring inflexibility, high cost and long deployment time. In the last decade, research on using wireless technologies as an alternative for structural health monitoring has been carried out but there is no systematic approach to investigating the challenging problems of wireless networks and building real-world systems.
Collaborating with researchers in civil engineering, Prof. Cao and his research team have developed a practical structural health monitoring system using cutting-edge intelligent wireless sensor networks. The system can be easily and flexibly deployed on various types of structures, and automatically form a self-organizing wireless network to fulfill the required monitoring functions. It has low cost while meets the practical civil engineering requirements and, therefore, can complement the wired systems to provide more coverage and real-time processing. Comparing with existing works on wireless structural health monitoring, Prof. Cao’s team has proposed computing-engineering co-design and developed various innovative yet practical techniques for energy saving, highly accurate damage detection, and fault-tolerance. The techniques include vibration-triggered wakeup, high accuracy wireless synchronization, distributed in-network processing, and automatic detection and recovery of faulty nodes. The prototype system has been tested in many real infrastructures, including the CCTV tower in Beijing and Hedong Bridge in Guangzhou, and demonstrated its high performance. The technology developed can also be used to monitor the safety condition of old tenement buildings and structures that might be damaged in a disaster.
The research was supported by GRF, Innovation and Technology Fund (ITF) and HK PolyU’s Niche Area Fund. The output of the research has made high impact not only on academic value in technology development of wireless sensor networks but also on its practical use and application to public safety. The research has won “Hong Kong ICT Awards 2013 - Special Mention Award and the Certificate of Merit”, the ‘Best Paper Award’ at IEEE WCNC 2011 and ISSNIP-2009, and produced over 15 research publications in top journals and conferences. It also attracts attention and requests for collaboration from industries.
With the ubiquitous use of various types of digital imaging devices in the current e-world, there is a vast and increasing proliferation of visual data, for example, in websites of YouTube, Facebook, Google, Flickr, and via networked TV, etc. Due to the limitation of low-end consumer electronics’ imaging sensors (e.g., smartphone cameras) and limited network bandwidth, the resulting image and video streams are often noise corrupted, blurred, have low resolution and artifacts, etc. It is thus of high importance to develop new technologies to enhance the visual quality of image/video streams to meet the increasingly higher requirements from users.
The research team led by Prof. Zhang has been working on image/video quality enhancement for a long time. Different from traditional methods which often work on the pixel level and enhance a pixel by using its neighboring pixels, Prof. Zhang’s team proposed to work on the patch level for image enhancement. The lift from pixel level to patch level greatly improves the local image structure preservation. The approach can go further by clustering patches into groups and performing group based enhancement to exploit the correlation between similar patches. The algorithms they developed achieve cleaner images, higher resolution, and sharper edges and textures. With these techniques, Prof. Zhang’s team has developed a real-time video quality enhancement software system, which can help users experience much better quality of IPTV programs without increasing their network bandwidth.
This research was supported by RGC GRF grants and ITC ITF grant. The research output has won the Best Paper Award in SPIE VCIP 2010 conference, and produced a highly cited paper published in IEEE Trans. on Image Processing in 2011 (cited for 230 times so far and ranked as the “hot papers” -- the top 0.1% cited papers in 2 years since its publication by Web of Sciences). Their “Digital Image/ Video Quality Enhancement System” was awarded the Silver Medal in the iENA exhibition held at Nuremberg Germany in 2010 with core technology patented in USA. The achievements of Prof. Zhang on image/video quality enhancement have also been covered by many local media such as Sing Tao Daily, Oriental Daily News, Ta Kung Pao, and Hong Kong Entrepreneur.
The Internet has already become an indispensable infrastructure for social interactions, business and government operations, world economy, and many good causes. Almost all popular apps rely on Internet connection for computation offloading, storage, information access, communication, and collection of user behaviour. However, the Internet is intrinsically unreliable, susceptible to various types of performance degradation and attacks. However, due to the highly distributed architecture of the Internet, identifying the root cause promptly and effectively for an observed performance problem - an open research problem - is already a daunting task.
The research team has been active for the last six years in developing active measurement methods and measurement platforms for measuring and monitoring Internet performance. The active measurement methods enable an ISP operator to measure the network performance perceived by their customers and a home user to gauge the actual network performance provided by her ISP. By coordinating the measurement from multiple users, we can also improve the network performance by changing to a better network path, revealing the cause of performance degradation, and detecting possible network attacks.
The research outputs were published in the top conferences in the field, such as ACM CoNEXT, ACM/USENIX IMC, and USENIX ATC, and in top journals such as the IEEE JSAC. They have a US patent on one of their new measurement methods to be issued soon. They are providing network monitoring services to the local universities by performing around-the-clock network monitoring for the Hong Kong Academic and Research Network (HARNET) and to a Hong Kong government department for monitoring local network infrastructure (due to the NDA, the project details cannot be revealed.)
This research has attracted close to $7 million of funding from three consecutive ITF grants (two tier-3s and one tier-2) and more than $2 million of consultancy fee (from two projects: HARNET and a government department) for the last six years. The HARNET project has been ongoing since 2009, and it has helped the local universities discover and diagnose various kinds of performance problems.
