Hardware Development of Brain Computer and its Application to Visual Information Processing

Features and Uniqueness
  • Toward realizing a high-performance neuromorphic vision processing system, our research involves a neural network model of spatial perception based on motion-stereo vision and its implementation on a very large scale integrated circuit (VLSI). Important parameters for spatial perception such as time-to-contact, orientation, and the shortest distance to plane surfaces of a target object are recognized by using the model without knowing moving direction and velocity of the object. We have developed a VLSI capable of performing visual perception based on the model with low power consumption, reduced less than 1/100 compared with a commercial CPU.
Practical Application

Targeted applications include a drone flying autonomously equipped with collision avoidance system by visual perception. Targeted industry includes infrastructure inspection, agriculture, logistics, and so on.



Research Institute of Electrical Communication

Shigeo Sato, Professor
Ph.D (Information Science)