In this modern social, people get used to a lot of smart devices, including smart phone and wearable devices. Here, we try to apply advanced information processing technology in those devices, to make people' life more convenience.
We propose a smartphone-based Driving Data Recorder (DDR). The conventional DDRs are standalone devices with multiple sensors and record many useless data or lose important information. On the other hand, the widely used smartphones already have the hardware conditions to replace the conventional DDR products. Therefore, we propose to develop the intelligent DDR using smartphones, which have two functions: motion sensor-based speedometer and vision sensor-based scene understanding.
For autonomous driving, ITS and smart-phone applications, it is significant to know the accurate self location. We focus on self-localization in urban canyons and are aiming to establish self-localization methods with high accuracy utilizing GNSS such as GPS and on-board/smart-phone sensors such as an inertial sensor.
For autonomous driving and Advanced Driving Assistant Systems (ADAS), it is essential to maintain safety for pedestrian and bicyclist. We approach various layers from pedestrian detection to behavior recognition for this theme.
3D maps are indispensable for self-localization technology such as camera and laser based ones and our 3D-GNSS. Accuracy of the maps affects on results of the self-localization. We are working on 3D map generation and correction methods.
Recently, there has been an increased demand for understanding the behavior and position of people within large-scale indoor facilities, such as customers within a department store. Current behavior and positioning analysis systems, however, require many sensing devices distributed across the area (e.g. cameras, Wi-Fi access points, and Bluetooth low-energy beacons) and can therefore be expensive. In this research, we propose an low-cost system to determine a person's activity and indoor positioning using only smartphone sensors (accelerometer and gyroscope).
Conventionally, retailers hire human labors or consulting companies to watch the surveillance video and extract useful marketing information. The video analyzers can assess how interested in the merchandise the customers are by observing their behaviors. However, watching the huge amounts of surveillance video is a very tedious and time-consuming work. The aim of our research is to automatically recognize typical customer behaviors in retail stores and estimate the interest level.
For the details and other research topics, please check Research page
Our laboratory is at Institute of Industrial Science in Komaba Research Campus. It belongs to Department of Information and Communication Engineering, Graduate School of Information Science and Technology, and Emerging Design and Informatics Course, Graduate School of Interdisciplinary Information Studies. We consist of 8 master course students, 5 doctoral course students and 2 researchers (from 2016 April) including a lot of foreign members, especially Chinese. We have open campus event once per year. You can also visit our laboratory anytime. Please contact us.
>>LastUpdate : Mar. 2nd, 2016
Mar. 2nd, 2016
Update research results in 2015 and other information.
Jan. 21th, 2015
Updated publications
Jan. 21st, 2014
Updated details of research topics, "Traffic signal control"
Nov. 25th, 2013
Updated details of research topics, "GNSS error correction in urban canyons"
"Activity recognition using mobile devices" and "Human pose recognition for retail surveillance".
Oct. 29th, 2013
Updated overview of researchs
Aug. 27th, 2013
Updated publications
Apr. 23th, 2013
Added "Committees" in the menu