The Study Of ECG Feature Extraction Method Based On Wavelet (Packet) Transformation

Affective Computing is one of the main technologies to achieve the high-level Human-Computer interaction,and Emotion recognition is the focus of affective computing research now.The objects of research in emotion recognition include facial expressions,speech sound,body posture,physiological signals and so on.ECG is the main object in the emotion recognition based on physiological signals.In the study of emotion recognition based on ECG,the main method is to obtain effective features after preprocessing ECG,and then classify emotion status.Therefore,extracting useful ECG features,analyzing the trends of changes in the features,identifying feature subset that can reflect emotional state changes are the basis to improve emotional recognition.In this paper,the study of emotion recognition based on the ECG was carried out,including the following two jobs:1.Because of lacking of ECG data used to analyze emotion,we carried on ECG collecting for the later studies.2.In the text,wavelet(packet) transform was used in ECG feature extraction for reducing the number of ECG false waveforms and improving the emotional recognition.The main method is:First,optimal mother wavelet is used to remove noise.Second,P waves,QRS waves,T waves were detected with the help of wavelet transform.Third,the energy of each wave was calculated and analyzed.Fourth,Features was extracted from energy and used as the input of classifier to verify the validity.This method was used to process ECG in Augsburg,comparing the recognition with their four physiological signals,has closing quote,and better recognition for joy with 10%higher.When the method was used to process our collecting ECG to extract features,which classified by Fisher,the recognition for joy and sadness is 80%and 80%.Test results show that ECG features extracted from two different sources can reflect changes in emotional states and the use of wavelet transform benefitted the feature extraction from ECG P-QRS-T waves,which can improve the emotional recognition rate based on the ECG,so the method of feature extraction in this paper is effective.