Recognition Of Facial Attraction Based On Photoplethysmography And Electromyography Signal

Facial attractiveness refers to a positive and pleasant emotional experience induced by the appearance of the target person,which can prompt others to want to get closer to the target person.Relevant studies have shown that facial attractiveness always has certain advantages when choosing a partner,looking for a job,or seeking help.It can be seen that it is of great significance to be able to accurately identify whether facial attractiveness is produced.Physiological signals have been proved by many studies to contain a lot of information about emotions,and have the features of not being easily affected by the subjective influence of the characters themselves.However,no researchers have been found to discuss facial attractiveness from the perspective of physiological signals.Therefore,this paper chooses to obtain two kinds of physiological signals that are convenient and effective,including Photoplethysmography(PPG)and Electromyography(EMG),and proposes a research on facial attractiveness recognition based on PPG signal and EMG signal.A model based on sequential backward floating selection algorithm(SBFS)combined with support vector machine(SVM),linear discriminant analysis(LDA)and XGBoost is used to identify whether the appearance of the target person has facial attractiveness.The main research work and corresponding achievements of this paper are as follows:1.Through the designed experimental paradigm of facial attractiveness induction,a database of physiological signals related to facial attractiveness was established.This paper designs an experimental paradigm that induces the emotion of subjects’ facial attractiveness through the appearance pictures of the characters,and selects 480(240 male pictures and 240 female pictures)portrait pictures with high,medium,and low facial attractiveness.Recruited 46(22 males and 24 females,age range: 17-27 years old,mean:21,standard deviation: 2.36)physically and mentally healthy college students to participate in the experiment,and obtain their PPG signals and signals during the experiment.EMG signal data successfully induced facial attractiveness.2.A recognition method of facial attractiveness based on the fusion of the physiological features of the PPG signal and the EMG signal is proposed.The best recognition classifier is obtained by comparing the recognition performance of three different classifiers.Firstly,the time domain,frequency domain and nonlinear features are extracted from the preprocessed PPG signal,and the time domain and frequency domain features are extracted from the EMG signal.Then the physiological features of PPG signal and EMG signal are fused and input into the classifier model,and the classification is performed through 10-fold cross-validation.The paper uses three classifier models,including support vector machine,XGBoost,and linear discriminant analysis.The comparison of the classification results shows that the XGBoost classifier has the highest recognition accuracy rate of 68.04%.3.Based on the recognition method 2,a facial attractiveness recognition model based on the combination of SBFS feature selection algorithm and classifier is proposed.First,by using the SBFS feature selection algorithm,the best feature subset in the fusion feature set is selected,and then input into the classifier model,and identification and classification are performed through 10-fold cross-validation.The classification results show that when the sequential backward floating selection algorithm is combined with the XGBoost classifier,the best emotion recognition accuracy rate is 71.53%,which is3.49% higher than the recognition method 2.Therefore,the facial attractiveness recognition method combining the sequential backward floating selection algorithm and the classifier proposed in the paper can effectively identify whether facial attractiveness is generated.At the same time,PPG signal and EMG signal are two physiological signals that can be applied to wearable devices,which have huge commercial value.In the future,it is hoped that the algorithm will be integrated into wearable devices,which has important reference significance for social activities such as partner selection and job search.