Emotional Analysis System Based On HRV And DSV Of Heart Sound Signals

Emotional recognition based on physiological signals is a relatively new and fast growing area.A variety of physiological signals have been successfully applied to emotion recognition.As an important physiological signal,heart sound directly reflects the beating of the heart.The beating of the heart is closely related to the emotions,so heart sound can reflect emotional changes.But there is no precedent for emotion recognition based on heart sound.The topic creatively applies the heart sound to emotion recognition,and constructs an emotion recognition system based on HRV and DSV of heart sound signals.Firstly,introduce the independent designed experiment platform and an emotional heart sound acquisition scheme was formulated.Then an initial emotional heart sound database was set up.A group subjects were administered with visual/audio stimuli while heart sound signals were continuously recorded with the shoulder belt heart sound-sampling device.The stimuli were grouped into four different kinds and the ECG signals were recorded simultaneously as comparison.Secondly,design a heart sound preselecting and segmentation algorithm based on template selection to automatically delete heart sound affected by strong noise.This algorithm can also accurately segment heart sound to calculating HRV and DSV.The algorithm mainly includes:(1)Some parameters of the compound heart sound generator based on Chaotic characteristic are optimized,and a maturity heart sound template library is established;(2)Preselecting heart sound to avoid meaningless segmenting;(3)Accurately segment preselected heart sound.The algorithm can effectively maintain the emotional heart sound database.Only the heart sound signals with a segmentation accuracy of 100% can eventually remain in it.Third,according to the characteristics of the heart sound signals,two emotion evaluation indexs based on heart sound were proposed——HRV(heart rate variability,HRV)and DSV(the ratio of diastolic to systolic duration variability,DSV).A set of emotion recognition index system of heart sound was designed to objectly evaluate this two indexs.The evaluation process includes:(1)Feature extraction on HRV of heart sound?DSV of heart sound and HRV of ECG;(2)GA algorithm is used to select feature and calculate the weight index of linear,nonlinear and graphic features;(3)According to the weight index to select the optimal feature subset of emotion and use the subset for recognition.Experimental results demonstrated that:(1)Both HRV and DSV of heart sound are effective emotional evaluation index as HRV of ECG;(2)Three kinds of emotional evaluation indexes are used to recognize emotions of different Valence?Arousal or Valence-Arousal integrated,HRV of Heart sound has the best recognition results with an average accuracy of 94.7917%?90.625%?80.2083%;(3)Combining the features of HRV and DSV together to recognize emotions of different Valence?Arousal or Valence-Arousal integrated,the average accuracy up to 94.7917%?90.625%?80.2083%.It shows that the presentation of DSV has great significance.Thus,the emotion recognition method based on heart sounds has obvious practical value in human emotion recognition.