针对脑电信号(EEG)运动想象分类过程中弱相关特征量影响分类准确度的问题,提出一种筛选方法,该方法是基于α波和主成分分析(PCA)算法的.基于脑机接口(BCI)系统,通过听觉诱发刺激产生向左和向右两种运动想象任务对应的脑电信号,并对其做小波包分解处理,然后进行脑电α频段信号的重构,从而提取出α波形并对其进行统计特征提取.再结合PCA技术和支持向量机(SVM)方法,实现弱相关特征的剔除和特征分类.根据筛选后的数据进行分类,所得结果准确率更高,信号分类的准确度由90.1％提高至94.0％.

根据地学仪器专业的实践教学特点,将自制科研仪器核磁共振找水系统JLMRS应用于地学仪器认识实习.根据核磁共振找水系统的工作原理和使用方法设计野外实验,介绍了数据处理的基本流程和简单的反演成像算法.基于核磁共振找水系统的地学仪器认识实习使本科生能够了解到最前沿的科研仪器设备,开拓了本科生的视野,激发了本科生对地学仪器专业的学习热情,有利于研究型和创新型人才培养.

针对奇异值分解(SVD)去噪中有效信号重构阶次难以确定的问题,论文提出基于K-K(Kohonen-K-means)神经网络的单道SVD地震噪声联合压制方法.搭建K-K神经网络,对单道数据重构矩阵分解所得奇异值进行无监督聚类,以确定有效信号重构阶次.经验证,该方法针对浅地层地震数据,无须反复试验对比,依据数据本身即可实现强周期性噪声和随机噪声的同时压制,减少多依靠操作者经验的情况,同时加强非水平同相轴去噪效果.

This research is carried out re-designing a brain-computer interface(BCI) system based on motor imagery recognition through extracting features of Alpha wave in electroencephalography(EEG)signal during motor imagery process,using multi-feature classification method in order to increase the accuracy of classification.Aiming at the shortcomings such as low accuracy and timeconsuming when one feature is adopted in the classification process,methods including AR model,statistical characteristics extraction and frequency domain analysis,etc.are taken to extract various features of Alpha wave.BP neural network is used to classify features.The system is designed to identify motor imagery and through experimental verification,it has achieved expected effect with high classification accuracy.The research proves the feasibility of brain-computer interface system combining multi-feature integration with BP neural network.