Speaker:  Professor Fei  Qiao

     Beijing TsingHua University.

Topic :  Signal Processing Architectures and VLSI for Energy-Efficient Intelligent Visual Perception System.

Date :  13:30～15:00 Wednesday 4 May. 2016

Place : 105 Delta Building (台達館)

Host:   Prof. Ching-Te Chiu

Abstract:

Visual perception applications, such as tasks of image feature extraction and classifications, are more and more prevailing in future industrial and consumer fields, such as VR/AR, CNN, face detections and auto-driving system, etc. However, these data-intensive, computing-intensive and memory-intensive applications need much more processing ability and energy-efficiency, which would not be affordable by current integrated architectures and circuits. In the post Moore era, some new signal processing architectures and VLSI implementation methods would be explored to release the constraints of current “Cycle-by-Cycle” computing paradigm. Fortunately, most visual perception algorithms are inspired by human-being, which would illuminate some idea of also bio-inspired hardware implementations. In this talk, we would introduce a signal processing architecture of analog to information conversion, which would use the mathematic/physical features and/or circuit topologies performing ultra-fast and energy-efficient complex tasks in analog domain. More, in some processing still in digital domain, approximate computing methods would be adopted to reduce the unnecessary clock cycle, due to the error-tolerant characteristics of most visual perception systems. 

Biodata:

Fei Qiao, Ph. D., Associate Professor with Dept. of Electronic Engineering, Tsinghua University. Fei Qiao is currently the group leader of iVip (integrated Vision, intelligent perception) @ NICS Lab of Tsinghua University. His research interests are low power CMOS circuits design for multimedia sensor network, energy-efficient circuits and architectures for integrated intelligent visual perception, as well as heterogeneous application systems for smart cameras and smart image sensors. Personal Webpage: http://nics.ee.tsinghua.edu.cn/people/qiaofei/ 

All faculty and students are welcome to join the lecture