基于二维特征和CNN分析的无人机操控员情绪状态检测研究

The Emotional Status Testing of UAV Operator Based on the Two-dimensional Feature Maps and CNN Analysis

摘要:

为了实时检测无人机操控员的情绪状态,提出了一种基于二维特征和卷积神经网络(CNN)分析的无人机操控员情绪状态检测算法。针对脑电信号(EEG)中眼电伪迹干扰的问题,设计实现了一种基于二阶盲辨识(SOBI)的去除伪迹算法。针对其它模型检测率低的问题,通过微分熵特征(Differential Entropy, DE)提取、2-DMapping映射及稀疏运算将一维脑电信号转化为包含情感信息的二维特征图,并对脑电信号进行扩增处理,提出二维特征图与CNN相结合的方式,使得各通道的情感特征相互关联。利用CNN自动学习深层次特征的优势,深度挖掘二维特征图里的脑电情感信息,较好的实现了无人机操控员积极、中性以及消极三种情绪状态检测。

Abstract:

In order to detect the emotional state of the UAV operator in real time, a UAV operator emotional state detection algorithm analyzed based on the Two-dimensional Feature Maps and Convolutional Neural Network(CNN). Aiming at the problem of the interference comes from ocular artifacts in electroencephalogram signals(EEG), a removal algorithm of the Second Order Blinding Identification(SOBI) is designed. For the problems of low detection rates of other models, extraction of one-dimensional brain electrical signal into a two-dimensional special symbol with emotional information through the Differential Entropy (DE) extraction, 2-D Mapping mapping and sparse computing, and the electrical signal is converted into emotional information. The amplification treatment is performed, and the method of combining the Two-dimensional Feature Maps with CNN is proposed to make the emotional characteristics of each channel interconnected. Using CNN to automatically learn the advantages of deep-level characteristics, and deeply excavate the emotional information of the Electrical Electricity in the Two-dimensional Feature Maps, it has better realized the three emotional states of the UAV operator positive, neutrality and negative emotional state.