As the growth of innovating computer vision applications and the advancement of semiconductor process technologies, more and more visual recognition systems have been developed to assist people and make our lives better. Existing works for vision technologies focus on software algorithm development, like image/object recognition and self-learning techniques. These algorithms usually pursue high accuracy of recognition rate but are very time consuming. Therefore, these works cannot be practical solutions for portable devices with both high recognition ability and real-time performance. Further, the architecture should be designed with low power specification. Motivated by this, we aim to develop an intelligent visual recognition system with both high performance and efficiency. Our design can be applied to many popular vision applications, like wearable visual-aids, robot vision and vehicle drive-assistant systems.



52mW Full HD 160-Degree Object Viewpoint Recognition SoC with Visual Vocabulary Processor for Wearable Vision Applications [ Webpage ]

Yu-Chi Su, Keng-Yen Huang, Tse-Wei Chen, Yi-Min Tsai, Shao-Yi Chien, Liang-Gee Chen

2011 Symposium on VLSI Circuits (SOVC)

Intelligent Vehicle [ Webpage ]

Yi-Min Tsai, Chih-Chung Tsai, Keng-Yen Huang and Liang-Gee Chen

"Detection and Tracking System for Automotive Applications", in IEEE International Conference on Consumer Electronics (ICCE), Las Vegas, U.S.A, Jan. 2011.