基于小样本学习的精分患者脑电信号分析方法

Method for Analyzing EEG Signals of Schizophrenic Patients Based on Small-sample Learning

目前很多机器学习方法都建立在大量数据的需求上,但在实际的生产生活中,有时很难获得大批量的数据.本文提出将小样本学习技术应用于脑电信号分析上,以增强分类精度,提高分析性能.实验通过对采集到的脑电信号进行预处理,再将其通过预训练模型的卷积层和池化层输出的高维特征向量作为输入数据进行特征提取,最后用小样本学习训练模型,实现在数据集规模较小的情况下达到较好的分类或预测效果.该方法结合了卷积神经网络和元学习的思想,通过在少量标记数据上进行训练,实现了对未标记数据的快速适应.实验结果表明,该方法在小样本情况下具有更好的分类精度和泛化能力,相较于传统的机器学习方法,具有更高的应用价值,可以为精神分裂症疾病的诊断提供参考.

Currently many machine learning methods are built on the demand of large amount of data,but in real production life,sometimes it is difficult to obtain large amount of data.This paper pro-poses to apply the small sample learning technique on EEG signal analysis to enhance the classification accuracy and improve the analysis performance.The experiment is carried out by preprocessing the col-lected EEG signals,then using their high-dimensional feature vectors outputted by the convolutional and pooling layers of the pre-trained model as the input data for feature extraction,and finally training the model with small-sample learning to achieve better classification or prediction results in the case of a small dataset size.The method combines the ideas of convolutional neural network and meta-learning,and achieves fast adaptation to unlabeled data by training on a small amount of labeled data.The experimental results show that the method has better classification accuracy and generalization ability in the case of small samples,and compared with the traditional machine learning method,it has higher application value and can provide reference for the diagnosis of schizophrenia disease.