Emotional Recognition Of Respiratory Signals Based On EEMD And Information Entropy

Physiological signal is closely related to psychological state,and the change of psychological state will affect the expression of physiological signal.As the most intuitive expression of psychological state,emotion is bound to be reflected in physiological signals.In recent years,with the rapid development of the society and the acceleration of the pace of life,people's mental pressure is increasing,resulting in the occurrence of more and more frequent negative emotions.The long-term impact of negative emotions can cause symptoms such as mania and anxiety in mild cases,and even lead to extreme cases such as depression or suicide in severe cases.Therefore,it is of great social significance and clinical value to timely and accurately identify people's emotional state and carry out correct intervention and treatment.On the other hand,as the research progress of artificial intelligence and machine learning and its rapid application engineering,human-computer interaction ability,especially the importance of emotion recognition and emotional computing is becoming more and more significant,in clinical studies,social sciences and engineering research and application of the role of growing,emotion recognition and emotional computing research not only has important theoretical significance and has important application value.In recent years,emotional recognition based on physiological signals has gradually occupied the mainstream and attracted more and more attention.In emotion recognition based on respiration signal is lack of research literature,the results of too little output,respiratory signal feature library feature quantity is not enough rich,in this paper,on the basis of existing research in the field,puts forward the new method for respiratory signal emotional feature extraction,and introduced the polymerization empirical mode decomposition and modern signal processing methods,such as information entropy,the emotion recognition based on respiratory signal problems,the main research content is as follows:(1)Design a breathing signal collection experiment which stimulated the subjects'specific emotions and recorded the breathing data of 6 emotions including calm,fear,sadness,happiness,anger and disgust for 5 minutes.Then,the second order IIR peak filter is used to filter and de-noising the resampled respiratory signal,removing the interference of the high frequency noise of the respiratory signal,and obtaining the relatively pure signal for the subsequent research,laying a foundation for the extraction of the characteristics of the signal.(2)Base on the EEMD decomposition algorithm,the approximate entropy,sample entropy,fuzzy entropy and energy entropy of each IMF component after the decomposition of respiratory signals were calculated to form the EEMD information entropy feature set of respiratory signals.And according to the different emotions by EEMD decompose the respiration signal after the IMF components with original signal correlation coefficient to choice of feature set,finally choose the strong correlation between the original signal and the weight calculation of the corresponding four kinds of information entropy to build EEMD respiration signal matrix,information entropy for emotion recognition model of respiratory signal feature set.(3)The EEMD information entropy feature group of respiratory signals was input into SVM and random forest,and the emotional recognition model of respiratory signals was constructed by using tenfold cross test and grid optimization to determine the optimal algorithm parameters,with the highest one-to-one emotional recognition rate reaching 83.33%.The feasibility and effectiveness of the EEMD information entropy algorithm in breath signal emotion recognition are demonstrated.(4)The respiratory signal is commonly used in traditional frequency domain feature extracting and primitive respiratory signal sequence of three kinds of entropy features a total of 15 kinds of features of the traditional group,the EEMD group,the traditional features of information entropy and the features of the combination of the two groups after fusion group respectively for emotion recognition and matching recognition rate,EEMD group information entropy good classification results show that the method is applied to the superiority of respiratory signal emotion recognition research.