Research On Emotion Clssification Technology Based On Mixed Physiological Signal Processing

Emotion classification and recognition is the study of human emotions,and related technologies can make intelligent machines better assist people's work.Especially with the development of wearable devices,physiological signals can be collected dynamically in real time.It is possible to classify the emotion of dynamic physiological signals based on wearable devices.However,at present,the research of emotion classification based on physiological signals focuses on the research of emotion classification based on EEG and peripheral physiological signals under laboratory conditions,and fails to do the research of mixed physiological signals fusion between EEG and peripheral physiological signals,and also fails to do the research of emotion classification based on dynamic physiological signals fusion which is more suitable for the actual scene.At the same time,the characteristics of physiological signals are proposed The physiological mechanism of emotion was not considered.Based on these problems,this paper studies the feature extraction algorithm from the physiological mechanism of emotional EEG,and constructs a four category emotional classification model of mixed physiological signals and a four category emotional classification model of dynamic physiological signals based on wearable devices.In the research of emotion classification based on EEG signals,this paper extracts the spatiotemporal features from the physiological mechanism of emotional EEG,and analyzes four kinds of emotions.It is found that spatiotemporal features have the best recognition effect,and at the same time,it can reduce the misjudgment of other state emotions as high arousal high valence and high arousal low valence can also improve the recognition effect of low arousal low valence state with bad classification results under other features.In this study,an eye noise recognition algorithm based on the similarity of independent components is proposed.The recognition result is 3.33% higher than that of the traditional independent component analysis method.For the research of mixed physiological signal fusion,this paper proposes two feature layer fusion methods,that is,the feature layer fusion algorithm of multi-layer perceptron based on voting mechanism and the feature layer fusion algorithm of multi physiological signal deep belief network.Compared with the traditional feature layer fusion algorithm,it is found that the two feature layer fusion methods proposed in this paper can improve the recognition accuracy of HAHV itself,and can also reduce At the same time,it can improve the recognition accuracy of HALV.Based on this,this paper proposes an emotion classification scheme based on the dynamic physiological signals of wearable devices.Compared with the results of EEG and mixed physiological signals,the model based on mixed physiological signals fusion has the highest recognition accuracy(0.8333),and the model based on peripheral physiological signals has a lower recognition rate(0.8137)than the former model.However,this method is suitable for the application of emotional recognition in the collection of dynamic physiological signals by wearable devices.