BP神经网络在脑电图信号预测中的应用

Application of BP neural network for forecast of EEG signal

由于EEG和BEAM的无创、低成本、能反映大脑病变部位的机能变化等特点,为了预测大脑疾病发生的可能性。设计了一个3层BP网络,并建立了相应的数学模型,以所记录数据中的一部分作为样本,利用Matlab的神经网络工具箱中的算法和函数,对上述神经网络进行训练、仿真、并预测另一部分数据,同时作出待预测数据和预测数据的脑电地形图,进行比较,验证了BP网络模型的有效性。其算法可开发出程序模块,集成到相关脑电图仪的软件系统中。

Because of the specialties that EEG and BEAM can harmlessly and low-cost reflect the functional changes at the ill part of the brain, in order to forecast the feasibility of the brain disease, a three-ply BP network is designed and the corresponding mathematical model is built. After the network is trained by taking one part of recorded EEG datum as samples and using the algorithm and functions in the Matlab toolbox, the simulation and forecast of the otherpart of the EEG datum are made. The BEAMs of the original and forecasted datum are given for comparison. The validity of the model of BP neural network is shown. A program module is developed using the algorithm and integrated into a software system of electroencephalograph.