Responses from two firing patterns in inferior colliculus neurons to stimulation of the lateral lemniscus dorsal nucleus

The γ-aminobutyric acid neurons（GABAergic neurons） in the inferior colliculus are classified into various patterns based on their intrinsic electrical properties to a constant current injection. Although this classification is associated with physiological function, the exact role for neurons with various firing patterns in acoustic processing remains poorly understood. In the present study, we analyzed characteristics of inferior colliculus neurons in vitro, and recorded responses to stimulation of the dorsal nucleus of the lateral lemniscus using the wholecell patch clamp technique. Seven inferior colliculus neurons were tested and were classified into two firing patterns： sustained-regular（n = 4） and sustained-adapting firing patterns（n = 3）. The majority of inferior colliculus neurons exhibited slight changes in response to stimulation and bicuculline. The responses of one neuron with a sustained-adapting firing pattern were suppressed after stimulation, but recovered to normal levels following application of the γ-aminobutyric acid receptor antagonist. One neuron with a sustained-regular pattern showed suppressed stimulation responses, which were not affected by bicuculline. Results suggest that GABAergic neurons in the inferior colliculus exhibit sustained-regular or sustained-adapting firing patterns. Additionally, GABAergic projections from the dorsal nucleus of the lateral lemniscus to the inferior colliculus are associated with sound localization. The different neuronal responses of various firing patterns suggest a role in sound localization. A better understanding of these mechanisms and functions will provide better clinical treatment paradigms for hearing deficiencies.

强混叠模式下基于神经元捕获/抑制原理的分类器设计

目前的许多分类器设计方法,如多层感知器网络(MLP)、支持向量机(SVM)、相关向量机(RVM)、径向基函数网络(RBF)等,实际是非线性映射加线性分类的方法,即将输入空间的非线性可分问题经非线性映射到另一空间,在那一空间实现线性分类.本文则开拓性的运用脉冲耦合神经网络神经元的点火捕获的思想,提出了一种基于耦合神经元点火捕获/抑制特性的分类器设计方法,使一类样本对应神经元总是较其它类样本对应神经元先点火以实现对样本的有效分类.所设计的分类器可实现对样本空间中任意复杂分布训练样本的非线性稳健分类,特别是有效实现复杂混叠模式的模式稳健分类,大量复杂混叠模式分类问题的仿真实验验证了本文方法的有效性和可行性,并应用于微波暗室实测一维距离像数据的自动目标识别中.

基于差别特征的纹理图像识别研究

文中提出了一种基于PCNN的捕获率图的纹理图像差别特征提取的新方法。通过对PCNN的基本特性分析 ,指出PCNN的捕获率图不仅包含了全局图像灰度分布的统计特征 ,还包含了局部图像灰度的空间分布的相关统计特征 ,即相邻象素之间的几何信息。这恰是纹理图像的个性特征所在。实验结果证明 ,该方法具有良好的鲁棒性。获得这些纹理图像的特征旨在度量纹理图像的相似性或从图像中抽取纹理区域。

一种新的纹理图像特征提取的方法

灰度图像的纹理反映了一个区域中像素灰度级及其局部变化的空间分布属性 ,用图像的局部统计特征能较好地刻画不同纹理的差异 .基于这种思想 ,本文中首次提出了一种基于 PCNN点火序列图的纹理图像特征提取的新方法 .通过对 PCNN的运行行为和基本特性的分析 ,指出 PCNN的点火时刻序列图不仅包含了局部图像的灰度分布信息 ,更重要的是还包含了图像中相邻像素之间的几何信息 ,这恰是纹理图像的个性特征所在 .最后给出了部分仿真实验的结果 。