Algorithm Of Psychological Stress State Analysis Based On SEMG

Psychological stress is a kind of emotional and physiological response whenindividual is faced up with a situation that requirements exceed the coping resources. Itdescribes one’s cognitive sense under mental or emotional stress. Constant stressfulconditions will affect one’s health and cause a series of pathological and physiologicalrisks.Physiological measurement is an effective evaluation method that detecting stresswith minimal discomfort for subjects.This study designed a stress stimulation scheme under the laboratory environmentusing surface electromyographic signals as the study object and the graduating students asthe subject. Imitate the real situations under the condition of laboratory, join a variety ofpsychological stress factors such as the well known high intensity noise and the low lightillumination. Different stressors were selected to stimulate the participants in thecontinuous four days to avoid the “adaptability” problem and improve the feasibility of thelaboratory stress induce project.With the influence of many external factors such as environmental changes,psychological state of the subjects and the different perception to different stimuli, thereexists a varying degree of diversity between samples and would have an impact on theevaluation affect. The approach to reduce individual differences is one of the hottest issuesin recent years and also the emphatically discussing issue in this research.The Support Vector Machine is used as the classifier in this paper. Two improvedalgorithms were proposed aiming at the individual differences, that is, the training sampleselecting and classification weight changing Support Vector Machine algorithm and thesupport vector selecting Support Vector Machine algorithm. The correlation betweensamples is considered form different perspective. Sample information was fed into themachine training process to reduce the negative effect in stress evaluation caused by thediversity problem.The classification accuracy of the training sample selecting and classification weightchanging Support Vector Machine algorithm has improved from75.33%to82.00%, and the algorithm running time has reduced from1365.3seconds to436.3seconds. While theclassification accuracy of the support vector selecting Support Vector Machine algorithmhas improved to82.67%and the algorithm running time has reduced to523.5seconds.Use Augsburg university stress data as a contrast in order to further verify the improvedalgorithm performance. Testing results show that the improved algorithm has a betterperformance than the normal algorithm either on accuracy or classification time.Experiment results showed that the improved algorithm can effectively address theinfluences of individual differences during stress state assessment and decrease thecomputation complexity of the classifier at the same time by improving the SVMclassification algorithm and adding the sample information to the pattern recognitionprocess.