摘要
为了提高动作识别的速度和准确性,本文研究了从初级视觉皮层V1中提取特征表示人体动作的问题,提出了采用integrate and fire(I&F)脉冲神经元模拟V1阶段神经元的方法。通过对脉冲输出进行分析,取脉冲序列平均发放率的熵,作为表征人体动作的特征向量,送入分类器进行分类。经过在Weizmann数据库下的测试,试验结果表明,本文的方法比Esco-bar[1]的方法更加有效.
In order to improve the accuracy of human action recognition and accelerate the recognition speed,this paper investigates the question of extracting feature from visual cortex for representing human actions,and proposes that integrate and fire neuron model was used to simulate the V1 neurons.According to the characteristic of spiking output,the entropy of mean firing rate of each neuron formed a feature vector,and Support Vector Machine（SVM） was used to classify different act.We compared our results with Escobar model.on the Weizmann action database.As a conclusion,we convinces that this method is more effective than the model of Escobar for human action recognition.
出处
《现代科学仪器》
2012年第2期29-32,36,共5页
Modern Scientific Instruments
基金
国家自然科学基金项目(60972158)
