摘要
为了满足神经网络实现的实时性要求,介绍了一种基于FPGA的神经网络可重构实现方法。首先在软件中利用改进的BP神经网络算法得到最优权值,然后在改进了激励函数逼近方法的基础上,用FPGA实现了羊绒近红外光谱模型的辨识。通过仿真实验得知,该方法有较好的辨识精度和速度,是一种有效的硬件羊绒近红外光谱建模方法,为羊绒、羊毛鉴别的嵌入式实现奠定了基础。
In order to realize real-time requirement,an implementing method of reconfigurable Neural Network based on FPGA was in troduced.First the optimal weights ware gotten by using improved BP Neural Network algorithm in software,and then on the base of im proved approaching method of excitation function the identification of cashmere Near Infrared Spectroscopy model was realized by using FPGA.The simulation experiments demonstrate that the method has better identification accuracy and speed,and it is an effective method of cashmere Near Infrared Spectroscopy modeling based on hardware,and it will laid the foundation for embedded realization of cashmere and wool identification.
出处
《电脑知识与技术》
2012年第3X期2092-2095,共4页
Computer Knowledge and Technology
基金
北京市优秀人才培养资助项目(2009D005001000003)
北京市教委基金项目(KM201110012010)
北京服装学院科学研究项目资助
关键词
神经网络
FPGA
可重构
羊绒
近红外光谱
neural network
FPGA
reconfigurable
cashmere
near infrared spectroscopy
