基于递归定量分析的生理信号情感识别-《计算机技术与发展》

文章信息/Info

Title:: Physiological Signal Emotion Recognition Based on Recursive Quantitative Analysis

作者:: 李才隆¹; 叶宁¹; 2; 黄海平¹; 2; 王汝传¹; 2; 1. 南京邮电大学计算机学院、软件学院、网络空间安全学院,江苏南京 210000; 2. 南京邮电大学江苏省无线传感网高技术重点实验室,江苏南京 210000

Author(s):: LI Cai-long¹; YE Ning¹; 2; HUANG Hai-ping¹; 2; WANG Ru-chuan¹; 2; 1.School of Computer Science,Software and Cyberspace Security,Nanjing University of Posts and Telecommunications,Nanjing 210000,China; 2.Jiangsu Provincial Key Laboratory of Wireless Sensor Networks,Nanjing University of Posts and Telecommunications,Nanji

Keywords:: recurrence plot; recurrence quantification analysis; statistical characteristics; feature extraction; emotion recognition

摘要:: 基于常规统计的人体生理信号特征提取应用广泛,然而基于常规统计特征的方法在分类识别效果上并不理想。为了解决这类问题,在研究人体生理信号的基础上,提出了一种基于递归图和递归定量分析相结合的方法,提取了生理信号在递归图中的递归率、确定率、对角线结构长度等特征。采用神经网络(neural network,NN)、K 最近邻(K nearest neighbor,KNN)、朴素贝叶斯(naive Bayesian,NB)、决策树(decision tree,DT)算法进行情感识别。实验结果表明,递归图中的特征是一组非常有效的特征。相对于传统的统计特征提取,非线性特征提取方法提取的特征更少,但是在情感分类识别效果上优于统计特征提取的方法。所采用的方法改进了传统特征提取数目庞大、效果不理想的问题,较好地实现了基于人体生理信号的情感识别。

Abstract:: The feature extraction of human physiological signals based on conventional statistics is widely used. However,the method based on the conventional statistical features is not ideal in classifying and distinguishing effect. In order to solve this kind of problem,we propose a method based on recursive graph and recursive quantitative analysis,and extract recurrence rate,determination rate and diagonal structure length of the physiological signal and so on. Neural network (NN),K nearest neighbor (KNN),naive Bayesian (NB) and decision tree (DT) are used for emotion recognition. The experiment shows that the feature in recursive graphs is an effective set of characteristics. Compared with traditional statistical feature extraction,nonlinear feature extraction has less features,but it is better than the method of statistical feature extraction in the effect of classification. The proposed method improves the problem of the large number of traditional feature extraction and unsatisfactory effect,which effectively solves the emotional recognition of human physiological signals.