基于验证集辅助的脑电信号包裹式降维

Enveloped Dimensionality Reduction of Electroencephalogram Signals Based on Verification Set

当今时代信息技术的高速发展促使人们对人机交互领域投以更多的目光,随时监测操作者脑力负荷情况并依此对操作者的任务工作量进行调整,在当下有着重要意义。有研究表明,脑电信号功率谱密度对于脑力负荷分类任务较为适用,但脑电特征维数较高,极易出现维度灾难。目前机器学习中降维方面应用最广泛的算法为主成分分析(principal component analysis,PCA),针对主成分分析在脑电信号分类上的不适应性和支持向量机(support vector machine,SVM)对特征间关系的敏感性,提出了基于PCA-SVM与逐阶枚举法的包裹式降维方法,在特征工程阶段引入固定验证集概念辅助包裹式降维,以验证集精度为指标调整特征工程方案,以此提高数据降维后的可分性。由于引入了监督学习概念,实验结果表明,基于PCA-SVM与逐阶枚举法降维过后的数据分类精度要普遍高于只依靠传统PCA的降维方式,以此为高维生物电数据降维提供了新思路。

Nowadays,the rapid development of information technology prompts people to pay more attention to the field of human-computer interaction.It is of great significance to monitor the mental load of the operator at any time and adjust the workload of the operator according to this.Studies have shown that EEG power spectrum density is suitable for mental load classification tasks,but the EEG(Electroencephalogram) characteristic dimension is high,which is prone to dimensional disaster.At present,principal component analysis(PCA) is the most widely used algorithm for dimension reduction in machine learning.In view of the inadaptability of PCA in EEG classification and the sensitivity of support vector machine(SVM) to the relationship between features,a wrapper dimension reduction method was proposed based on PCA-SVM and step by step enumeration method.In the feature engineering stage,the concept of fixed verification set was introduced to assist packaged dimension reduction,and the feature engineering scheme was adjusted by ta-king the accuracy of verification set as the index,so as to improve the separability of data after dimension reduction.Due to the introduction of the concept of supervised learning,the experimental results show that the data classification accuracy after dimensionality reduction based on PCA-SVM and step by step enumeration method is generally higher than that only relying on the traditional PCA dimensionality reduction,which provides a new idea for the dimensionality reduction of high-dimensional bioelectrical data.