Font Size: a A A

Study On Emotion Recognition Methods Of Stress State Based On Physiological Signals

Posted on:2014-02-04Degree:MasterType:Thesis
Country:ChinaCandidate:H Y SunFull Text:PDF
GTID:2248330392961185Subject:Biomedical engineering
Abstract/Summary:PDF Full Text Request
Fast-paced modern society and the fierce competition make eachperson subject various stress comes from work, study, life and so on,especially for athletes who have to face up to all kinds of competitions.Stress can induce different kinds of emotions such as Excitement, annoy orterror. Also, stress with different levels can induce different degrees of aspecific emotion. Physiological signals are generated by body’s internalorgans and they change with the emotions. The physiological signalsreflect the real emotional state objectively at the time.This paper is based on the process of the emotion recognition byphysiological signals, and it designed experiments to induce differentemotions and three level tension emotions of the volunteers. We collectedthe physiological signals such as heart rate and skin conductor, and alsoextracted and selected features from them. Then we make classificationand identification from the data. This paper combined k-nearest neighboralgorithm and support vector method with the Particle SwarmOptimization method separately to optimize the features selection andclassification from the emotion related features of multi physiologicalsignals under stress state. The result shows that the classification accuracyof the data which eliminated the baseline is much better than the datawithout eliminating. The kNN method with PSO is little better than theSVM method with PSO. Besides, we also have a research on the featuresof physiological signals which related with emotions much more, andestablished an emotion recognition model based on many physiological signals under stress. At last, after little adjustment we applied theexperiment recognition model on the rowing athletes to recognize theemotions before the competition. The result shows that the recognitionaccuracy rates of the PSO-kNN and PSO-SVM method are all good andneck and neck. The technology of the emotion recognition based onphysiological signal can measure the mental stress state the people suffered,which provides objective basis and method to the evaluation andintervention of stress state and has a profound significance.
Keywords/Search Tags:Emotion recognition, Stress, Particle Swarm Optimization, k-Nearest Neighbor, Support Vector Machine
PDF Full Text Request
Related items