| With the rapid development of society,today’s human-computer interaction is no longer satisfied with the purely logical interaction of the mouse and keyboard to enable the machine to understand and accept human commands.We also hope that the machine can understand various emotional characteristics and respond accordingly like human beings,which requires adding emotion recognition function to the machine.Before,single-modal emotion recognition or fusion of single biosignal emotion recognition has made great progress.For example,single-modal skin electrical emotion recognition or fusion of skin electrical,pulse,brain and other emotional recognition has achieved significant results.Among them,text information is one of the most commonly used communication methods for human beings.However,it is natural to identify emotions through text information.It is subjective and often people can hide their emotions.The electrical signal of the skin as a physiological parameter is only distributed by the autonomic nervous system and the endocrine system,and is not subjectively controlled by humans,thus preventing artificial "falsification" and it is easier to collect than other physiological parameters.The bimodal emotion recognition proposed in this paper is to fuse text information with skin electrical signals.The two-modal fusion emotion recognition itself enriches the emotional information,and the different types of emotional signals from the source make up the same biological signal..Firstly,the text information is preprocessed,including specification coding,denoising,word segmentation and part-of-speech tagging,syntax analysis,and then feature extraction.Secondly,the sensory sensor is used to collect the skin electrical parameters of the experimenter,and the physiological signal database is established.Then the collected skin electrical parameters are denoised,28 statistical feature values are extracted,and the data is normalized.Then,by extracting and optimizing the emotional key feature parameters of skin electrical parameters and text information,the artificial neural network algorithm of support vector machine algorithm and genetic algorithm optimization is designed as a single modal emotion classifier.Finally,the sample data and database needed for the bimodal emotion recognition experiment are prepared.The Gaussian mixture model optimized by the sorting election algorithm is used to weight the fusion of the decision layer,and the test is completed and compared with two single-modal emotion recognition.Experiments show that the accuracy of emotion recognition is significantly improved. |
