摘要
针对脉冲神经网络(SNNs)在视觉颜色特征分类领域精度不高的问题,提出了一种基于新型RGB-HSV预处理模型的高精度脉冲神经网络。该脉冲神经网络融合了RGB颜色通道简单和HSV色彩空间直观的特点来提取聚类色彩特征,增强了网络的识别能力。此外,在Tempotron有监督学习基础上提出了一种结合权值动量的训练方式,该方式在计算当前权值更新量的同时保留一定程度的上次权值更新量,加快了网络权值的收敛速度,节省了仿真时间。仿真实验结果表明,所设计的脉冲神经网络的分类精度高达96.21%,且在6次训练迭代后精度仍可达84%左右。
A spiking neural network with high precision based on a new RGB-HSV preprocessing model is proposed to solve the problem of low accuracy of spiking neural networks in the field of the classification of visual color features.The proposed network extracts cluster color features by combining the features of the simplicity of RGB color channels and intuitive of HSV color space and enhances the recognition ability.A training method with weight momentum is also proposed based on Tempotron supervised learning rules.The method updates weights with new calculations while some of last weights is retained,so that the convergence of network weights is speeded up and the simulation time is saved.Experimental results show that the classification accuracy of the proposed network is up to 96.21%,and the accuracy reaches about 84%after 6 training iterations.
作者
苏亚丽
吴健行
惠维
张国和
SU Yali;WU Jianxing;HUI Wei;ZHANG Guohe(School of Mechanical Engineering,Xi’an Shiyou University,Xi’an 710065,China;School of Microelectronics,Xi’an Jiaotong University,Xi’an 710049,China;School of Computer Science and Technology,Xi’an Jiaotong University,Xi’an 710049,China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2019年第10期115-121,共7页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目（61772413）
关键词
脉冲神经网络
特征分类
颜色特征
网络权值
spiking neural networks
characteristic classification
color feature
network weight
