Key Problems Research On Experimental System Of Emotion Recognition Based On ECG Signals

With the deepening of "human-computer emotional interaction" research,it becomes more and more important for computers to understand human emotional states.Compared with the facial expression,voice intonation,gestures and text emotion recognition in emotion research,ECG signals are objective and difficult to conceal,which has an irreplaceable role and significance.Also in the stage of emotion research,the quality of ECG data especially affects the advancement of emotion recognition research.At present,experimental emotions are often used to obtain target emotional data,but the evoking materials in experimental emotions are mostly manually selected by the experimenters,and the data labels are calibrated based on the subject's self-perception.The accuracy of the data needs to be improved.Combined with the traditional emotion experiment induction,an experimental system platform for emotion recognition based on ECG signals is proposed.It solves the flexibility and convenience of designing the induction scheme,and based on the traditional data screening,combined with the PAM algorithm for emotion recognition to complete the effective data screening,instead of the subject's self-assessment questionnaire to complete the emotional data screening.In the stage of designing the evoking scheme,combining the traditional picture evoking scheme design,taking into account the progressive state of emotion.Starting from the self-tagging of emotion evoked by the picture,an automatic generation strategy of picture evoking scheme is proposed.The system provides the valence and arousal range of the target emotional picture.The titer of the plan is positioned between the median titer and the titer deviation.The degree of stimulation of the picture is divided into three segments of low-medium-high based on the deviation of arousal degree,and an induction scheme is generated according to the degree of stimulation to induce the experiment.In the analysis of the self-collected data to complete the program evaluation,the PAM algorithm based on information gain is used to realize emotion recognition.The information gain is introduced in the feature selection stage,so that in the actual application of the system,the contribution of each feature to the current sample category prediction is obtained by analyzing the sample features for feature selection.First design experiments to automatically generate "happiness" and "sadness" schemes for emotion induction,and obtain the emotional data of ECGsignals.Then extract HRV time-domain,frequency-domain and non-linear domain features for information gain analysis experiments to determine the information gain size of each feature and the information gain threshold suitable for the current emotion.The combination of features above the threshold is used to complete emotion recognition.The results prove that feature recommendation based on information gain combined with PAM algorithm is suitable for emotion recognition.Then realize the evaluation of the induction plan and the screening of effective emotional data.