Research Of Emotion Recognition Based On Cardiovascular Signal Features And Heart Rate Variability

Emotion recognition or emotion measurement has been applied widely,such as social science,computer engineering and medical science.In medical science,emotion is not only the response to stimulation,but also the symbol of health.In addition,emotion may affect people’s psychological and physical.Mood disorders can lead to cardiovascular diseases and make them worse.Emotion recognition can reduce the adverse impact of emotions on cardiovascular system at prevention stage of.And it can also play an important role in the treatment of cardiovascular diseases,which can promote the recovery of cardiovascular disease.The cardiovascular system is a closed pipe system including of heart and vascular.Analyzing of and heart rate variability is an important method for researching cardiovascular system.However,when researchers did study of emotion recognition with photoplethysmograph and heart rate variability,features extracting from is dispersive and from nonlinear analysis is not enough.Thus,this paper will extract cardiovascular function features from photoplethysmograph in time domain and frequency domain and apply permutation entropy and permutation min entropy to study heart rate variability during different emotional states.And the difference of features during emotional states and effect of emotional states on the cardiovascular system were explored.The main contents of this paper are as following:(1)PPGs and ECGs of sixty subjects during emotional states were recorded.Wavelet packet transform and minimum value fitting are used remove the noise in PPGs and ECGs.Threshold algorithm based on mathematical morphology was used to detected starting points,peak points of PPGs and R peaks of ECGs.And RR interval time series were generated.(2)Cardiovascular function features were extracted from PPGs’ time domain and frequency domain analysis.Analyzing of heart rate variability was done in time domain,frequency domain and nonlinear analysis.Permutation entropy and permutation min entropy were used to determine the complexity of HRV during emotional states.And equalities in time series may cause a deviation from true permutation entropy and permutation min entropy and how to solve this problem was explored.Finally,Emotion was quantified and cardiovascular system is characterized by above features.(3)According to,the differences in cardiovascular system between neutral and emotional states was analyzed by using features extracting from PPGs and HRV.Then,change of physiological response during emotional states was hypothesized based on features statistical analysis.To recognize emotions,four steps were to prepare.First feature matrix was built;second,decision tree and support vector machine were used as the basic learning machine;third,system error and difference were set as criterion;finally,deleting worst base learner was used to ensemble prune.The ten-fold crossover test results showed that ensemble learning system performs better than single decision tree and support vector machine.