Understanding of Human Perceptual-Alternation Phenomena and its Neural Network Model

在本研究中，我们探讨了人类在观察“鲁宾的脸和花瓶”等模糊图形时的知觉转换现象。利用M.Riani等人提出的多层神经网络模型来描述大脑对这一现象的高级信息处理过程，我们研究了感知交替的放电率对输入模式的依赖性。在我们的方法中，这种交替被解释为感知模式的神经元放电率。实验上，我们可以得到交替解释的知觉持续时间的Gamma分布。然后，我们把这种伽马分布看作是知觉交替的普遍特征。计算机仿真结果表明了该模型的有效性。特别是，我们已经显示了对应于注意力眼球运动的视觉输入的混沌行为，使得有可能获得比随机和/或1/f视觉行为更清晰的Gamma分布。我们还发现了混沌输入的一些其他效率，例如，如输入模式依赖的放电神经元的速率，并讨论了另一种模型，例如，混沌神经网络。

In this research, we dealt with the phenomena of human perceptual-alternation at the time one observes the ambiguous figures such as the "Rubin's face and vase". Using a primary model based on the multilayr neural network proposed by M.Riani et al. to represent the information processing in the higher level of brain for this phenomenon, we examined the input pattern-dependence of the firing-rate for the perceptual-alternation. In our approach, this alternation is interpreted as that of the rate of firing neurons for the pattern perceived. Experimentally, we could obtain the Gamma distribution of the perceptual duration of the alternating interpretations. We then regards this Gamma distribution as the universal character of the perceptual alternation. From our result of computer simulation we could show the efficiency of the model. In particular, we have shown the chaotic behavior of visual inputs corresponding to attention eye movement makes it possible to obtain clearer the Gamma distributions than those of the random and/or 1/f visual behavior. We also find some other efficiencies of the chaotic inputs, for example, such as the input pattern-dependence of the rate of firing neurons and discuss an other model, for example, the chaotic neural network.