多模式无线数据采集动态神经网集成融合算法

Neural networks integrated fusion algorithm for multi-mode wireless data sampling system

针对现有数据采集系统中采用有线方式或单一无线方式传输数据存在的问题,诸如:安装不灵活、稳定性差,提出了一种多模式无线数据采集系统的设计方案。新方案有机地结合了蓝牙、IEEE802.11b、HomeRF和nRF2401各自的特点,使得系统可以灵活、准确、有效和低成本地进行无线数据采集与传输,并提高了数据采集系统的抗干扰能力。在对数据的融合中,提出了一种动态神经网络集成算法。该算法采用动态聚类算法实现对各个体神经网络输出数据的集成,较为有效地克服了单个神经网络泛化能力差的缺点。

In order to solve problems such as inflexible installation and poor stability in existing data sampling system which adopts wire mode or single wireless mode to transmit data, a solution of multi-mode wireless data sampling system is proposed. The proposed solution organically unites the characteristics of Bluetooth, IEEE802.11b, HomeRF and nRF2401 to enable the system to flexibly, accurately, effectively and cheaply sample and transmit wireless data, and promotes the anti-jamming ability of the data sampling system. A dynamic neural networks integrated algorithm in data fusion is presented. The algorithm employes dynamic clustering algorithm to integrate the output data of each neural network and effectively improves generalization ability of single neural network.