论文部分内容阅读
运动感知是生物视觉系统的一个重要组成部分.动物对自然界的认识一大部分是通过运动感知.本文通过对同时具有方向和速率选择性的视觉皮层运动区神经元的侧抑制机制的模拟,提出了一种新的乘积抑制速度检测子(MultiplicativeInhibitoryVelocityDetector-MIVD),其时空联合参数K决定了检测子的最优速度,基于MIVD的响应幅度差特性,我们构造了两个多速度融合神经网络(一个简单神经网络和一个主动神经网络)来检测一维运动速度.实验结果表明具有反馈的主动MIVD神经网络能够给出较好的检测结果.
Motion perception is an important part of the biological vision system. Most of the animal’s understanding of nature is perceived through movement. In this paper, we propose a new Multiplicative Inhibitor Velocity Detector (MIVD) based on the simulation of the lateral inhibition mechanism of neurons in the visual cortex with both direction and velocity selectivity. The space-time joint parameter K determines the detector Based on the MIVD response amplitude difference characteristics, we constructed two multi-speed fusion neural networks (a simple neural network and an active neural network) to detect the velocity of one-dimensional motion. The experimental results show that the active MIVD neural network with feedback can give better detection results.