- vLecture: Harnessing Lecture Videos for Learning
- Daily Activity Exercise Stress Evaluation
- Distributed Wireless Communication System over Sora
- Remote Sensing Algorithm Platform in Windows Azure
- A Unique Reminder for Mobile Phone Users: WhozTalking
- Boosting Entity Relation and Relationship Extraction Using Renlifang
- Global Cancer Map: Large-Scale Meta-Analysis of Cancer Microarray Data
- Sight-Related Web Images Enrich Travel Notes with Textual Context and Image Understanding
- Wireless Sensor Network Facilitates Smart Healthcare and Home Environment
- Intelligent Wheelchair Improves Mobility for People with Disabilities
- Hardware Accelerator Enables Easier and Faster Computing
vLecture: Harnessing Lecture Videos for Learning
Xinbo Gao, professor at Xidian University in China, collaborated with Microsoft Research Asia to develop vLecture, a virtual learning community that provides an efficient content-based video searching engine and a platform for learning and sharing knowledge online. Users can easily find the clips of lecture videos by inputting keywords, collect videos, make notes, and share knowledge with other users. vLecture consists of nearly 10,000 lecture videos in the field of computer science and technology, more than 2,000 indexed XMLs, and more than 60 registered users. In the future, we expect vLecture to facilitate learning without time or location limitations; it will gradually evolve into a social network that greatly contributes to sharing and communications.
Daily Activity Exercise Stress Evaluation
Appropriate amounts of exercise stress play a significant role in maintaining human health. Monitoring is useful to track motor functions and health status, and to understand the suitable degree of exercise stress for the individual. Deyu Li, professor at the School of Biological Science and Medical Engineering, Beihang University in China, collaborated with Microsoft Research Asia to develop a wearable device that monitors the plantar pressure distribution and foot movement continuously and also collects and transmits data with non-interference. The device will be used to inform users about their health status and suggest appropriate exercise methods. Additionally, it can be used by patients to evaluate rehabilitative exercises and by senior citizens and individuals with disabilities to select appropriate rehabilitation measures and assistive walking tools. It is expected that this device will be cost-effective, low-interference, low-energy and portable.
Distributed Wireless Communication System over Sora
A simple distributed multiple-input-multiple-output (MIMO) demo system has been developed by Ming Zhao, associate professor, Tsinghua University, to demonstrate the potential capability of the Microsoft Research Software Radio (Sora) platform in full distributed wireless communication systems (DWCS). With the support of Microsoft Research Asia, Prof. Zhao used the Microsoft Research Software Radio Academic kits to build a promising distributed wireless access system architecture with distributed antennas, processors, and controlling. The system’s capacity can be expanded through dense frequency reuse and co-processing of signals while reducing the transmission power required. The successful use of the Sora platform in distributed MIMO systems is expected to help establish more expertise in general purpose processors (GPP) based software radio techniques in next-generation telecommunications.
Remote Sensing Algorithm Platform in Windows Azure
Beijing Normal University, in collaboration with Microsoft Research Asia, constructed an open and flexible research platform by using Windows Azure, Microsoft .NET, and Microsoft Silverlight. The platform is known as Leaf Area Index (LAI) and efficiently calculates the amount of plant life by using large satellite images. LAI provides a solution of mass remote-sensing data computing that uses cloud computing to facilitate the sharing and communication of scientific research achievements. The LAI inversion algorithm can be used in a wide range of applications, including agriculture, weather, and ecology. It is expected that the platform will be the first and the largest system running a remote-sensing algorithm in Windows Azure, providing useful services for scientists.
A Unique Reminder for Mobile Phone Users: WhozTalking
It is not uncommon for people to forget the name of an acquaintance in an impromptu conversation or certain talking points from a meeting. Time-based or location-based reminders are used, but these require users to pre-set in advance. Xiao Bai, associate professor from School of Computer Science and Engineering at Beihang University, in collaboration with Jacky Shen, lead researcher from Microsoft Research Asia, developed a practical reminder called WhozTalking for Windows Phone users. WhozTalking is built on advanced voice-based identification technology and addresses challenges such as the bootstrapping issue and voiceprint enrollment problems while supporting a speed and accuracy priority mode. These two modes are designed to meet practical requirements under various real-life situations.
Boosting Entity Relation and Relationship Extraction Using Renlifang
Seung-won Hwang, professor at Pohang University of Science and Technology in Korea, proposed a novel data-intensive technique for automatically translating named entities, such as the name of a person, location, or organization. Although named entities are frequently mentioned and queried on the Internet, translations are challenging since new names appear every day in news articles but translations often cannot be found in dictionaries. The quality of translation is also a bottleneck in machine-translation systems and cross-language information retrieval applications, making it important for products such as search engines or translation applications to mine high-quality translation pairs. For this purpose, Prof. Hwang collaborated with the natural language computing group at Microsoft Research Asia to develop a new holistic approach that uses not only name or context similarity, but also entity co-occurrences and relational similarity between entities. The quality of the resulting pairs was shown to be significantly higher than productized resources such as Google Translate. These results were published at the EMNLP 2010 (Empirical Methods in Natural Language Processing) conference and CIKM 2011 (International Conference on Information and Knowledge Management).
Global Cancer Map: Large-Scale Meta-Analysis of Cancer Microarray Data
The deregulation of pathways is casually linked to the development and progression of cancer, but most of the previous cancer microarray meta-analyses have focused on finding individual biomarker genes; not much attention has been paid to the relation between the pathways and the different types of cancers. Jaewoo Kang, professor at Korea University, was sponsored by Microsoft Research Asia to conduct a meta-analysis of cancer microarrays to reveal the pathway hotspots that are linked to cancer. The global cancer map is expected to greatly assist the bio-medical research community as it offers an intuitive, easy-to-operate, information-rich navigational cancer analysis tool. The tool can fuel cancer research on an unprecedented scale, by providing access to tens of thousands of cancer samples over many different types and species. Jaewoo Kang has written about this work in four papers that have been published in top-tier bioinformatics journals.
Sight-Related Web Images Enrich Travel Notes with Textual Context and Image Understanding
A web service to help travel planning is the result of research conducted by Prof. Pang Yanwei from Tianjin University, China, and Zhang Lei, Lead Researcher at Microsoft Research Asia. When a user enters a scenic spot query, the top 50 textual travelogues pop up. The service automatically enables a search of related images to match the travelogues so that users can better understand the text. The service is based on 100,000 travelogues and 2,450,000 images from the web, which enable the user to search up to 32,000 popular locations. The project integrates Microsoft Silverlight and Virtual Earth technologies and fully achieved its research milestones with publications at ACM MM 2009 and LBSN 2009.
Wireless Sensor Network Facilitates Smart Healthcare and Home Environment
The intelligent wheelchairInspired by the compound insect eye structure, the Microsoft Research Asia – Xi’an Jiaotong University Joint Lab in China deployed a wireless sensor node with a camera—called a visual node (or mosaic eye)—inside a building to enable navigation of an intelligent wheelchair. The wheelchair is equipped with a PDA navigation controller and can be guided by a distributed intelligent sensor network according to semantic commands. The system can be applied to residential or healthcare environments to help people with disabilities locate and move to desired destinations.
Intelligent Wheelchair Improves Mobility for People with Disabilities
Dr. Feng Zhao, Assistant Managing Director of Microsoft Research Asia, tries out the wheelchairProgress on an intelligent wheelchair with a manipulator is the result of research conducted by Microsoft Research Asia and Shanghai Jiaotong University’s joint lab. The wheelchair is integrated with a variety of functions including environment sensitivity, dynamic planning, and decision-making, among others. It dramatically increases the level of mobility for people who use wheelchairs. The manipulator for @home league is a 6 DOF one. It is 1 meter long when expanded. The payload can reach 15N. It has a simpler gripper, a 3 DOF wrist that makes it possible for the manipulator to grip, hold, and change the position of the desired objects. A camera is fixed at the wrist for searching and recognizing objects. The user can control the wheelchair and manipulator by joystick or voice commands. Read more...
Hardware Accelerator Enables Easier and Faster Computing
Brain network research benefits from hardware acceleratorsA recent collaborative project between Microsoft Research Asia and China’s Tsinghua University focuses on finding solutions to accelerate most of the machine learning algorithms as an application domain in order to provide an efficient hardware architecture and interface to the operating system. Using Microsoft Research Asia’s recent research on accelerated boards, the project enables easier and faster computing, which will be particularly helpful to researchers in scientific domains such as brain networks and environments with limited computing capability.