摘要
大脑中致力于运动信息处理的区域是初级视皮层(V1)和中颞区(MT).目前有关运动模式是在哪个区域完成的,存在不同的推测.迄今大多数关于动作识别的研究都是围绕MT阶段展开的.本文针对V1阶段获得的信息能否进行动作识别的问题展开研究,提出了模拟初级视皮层(V1)脉冲神经元的动作识别系统.该系统首先采用3D Gabor滤波器及其组合分别模拟初级视觉皮层中简单、复杂细胞的感受野,以此对视频图像进行处理,从而获取对运动速度和方向敏感的运动能量,并通过V1阶段的环绕抑制来增强运动能量和降低噪声的影响.其次,采用Integrate-and-fire脉冲神经元模型模拟初级视觉皮层的神经元,将获取的运动信息转换为神经元响应的脉冲链.最后,根据脉冲链平均发放率的特性提取运动特征向量,采用支持向量机(Support vector machine,SVM)作为分类器.在Weiziman数据库下进行测试,实验结果表明,V1阶段获得的信息可以进行动作的识别.
There are several theories which speculate on how and where pattern motion is computed from visual cortex or middle temporal area （V1/MT） dedicated to motion. So far, most researches in action recognition remain rooted in MT. This paper proposed a method of human action recognition by modeling the human V1 neurons for information obtained in V1 which could benefit action recognition. The method firstly simulates the classical receptive field （CRF） of simple and complex cells in the primary visual cortex with 3D Gabor filter and its combination to process the video sequence, in order to obtain the sport energy that is sensitive to the sport speed and direction. Meanwhile, it enhances the sport energy and reduces the influence of noise through surround inhibition in V1. Secondly, conductance-driven integrate and fire neuron model is used to simulate the primary visual cortex neuron, by which motion information is converted into spike train. Finally, the mean firing rate of spike train forms a feature vector that captures the characteristic of human actions in this video sequence. Using support vector machine （SVM）, the method is tested on the Weizmann action dataset. The obtained impressive results show that the information obtained in phase V1 could benefit action recognition.
出处
《自动化学报》
EI
CSCD
北大核心
2012年第12期1975-1984,共10页
Acta Automatica Sinica
基金
国家自然科学基金(60972158)
湖北省自然科学基金计划重点项目(2011CDA078)
中南民族大学中央高校基本科研业务费专项资金项目(CZQ12011)资助~~
