摘要
基于最新的工业应用双核处理器OMAP—L137设计了一种神经网络软测量模型在线训练平台。该平台以ARM与浮点DSP双核控制器为核心。针对神经网络隐含层节点数、训练中止条件、学习率等相关参数的确定进行了研究与测试,将优化的神经网络算法成功移植到嵌入式平台,支持神经网络软测量模型的在线训练。通过对相关数据集的测试,结果表明:系统具有高速、高精度等优良性能。
A soft measurement model online training platform is designed based on OMAP-L137 ,which is a new double-core processor for industry. By analyzing the hidden-layer nodes of neural network, stop criterion of training and learning rate, the optimized neural network（NN） is realized in the double-core （ARM and floating point DSP） platform that supports on-line training of soft measurement model. The testing of datasets demonstrates the platform has excellent performance of high-precision and high-speed.
出处
《传感器与微系统》
CSCD
北大核心
2010年第8期100-103,共4页
Transducer and Microsystem Technologies
基金
国家"863"计划资助项目(2009AA04Z154)
关键词
嵌入式系统
软测量
在线训练
神经网络
embedded system, soft measurement, online training, neural network（NN）
