Stress Detection Based On Wearable Device And The Application In Robot Service Cognition

With the popularization of service robot,the demand of intellectualization in household robot is growing.People hope the robot can detect user's emotional changes,understand user's inner emotional needs and provide more personalized service,like human beings rather than a cold machine.Although intelligent space technology started a new way for robot service cognition,service robot often ignores user emotional factors in autonomous service cognition,resulting in unsatisfying service.Therefore,the ability of emotion detection is the premise and guarantee of the service tasks planning on the humanization and intellectualization.In order to further enhance the ability of human-robot interaction and improve user experience degree of the service,an emotion detection method is heavily needed.Furthermore,a user emotion based autonomous service cognition method and personalized service selection strategy for robot should be designed.The main four parts of this paper are as follows:(1)A stress detection method is presented as a part of emotion detection system.Firstly,emotion induce experiment is carried out and physiological information is collected through built-in sensors in wearable device.Then filters are used to remove the interference while keeping the original information.Finally,several emotion-related features are extracted from the pre-processed physiological signals.(2)In order to eliminate unnecessary information in features,principal component analysis is used to reduce the dimension of high-dimensional feature matrix.The original feature matrix is summarized by linearly independent principal components which selected by reconstruction threshold.Support vector machine is then designed to detect stress.Finally,the standardization formula,kernel function,and super parameter optimization method are compared and selected.(3)Considering the user emotion is always neglected during autonomous service recognition,an emotion-space-time ontology model is built based on information of user emotion and space-time.The emotion-space-time rule base is encoded and then used for training BP neural network reasoner.The real-time updated information in intelligent space and trained neural network are matched to generate the robot services.Finally,user emotion is used as a feedback to adjust the user preference of each subclass service,the personalized service selection is achieved.(4)The experiment and analysis are carried out on the stress detection method and the autonomous service cognition strategy.The experimental results show that the physiological information of wearable device can effectively identify the different degrees of stress;the service autonomous cognition method can realize the user emotion based service planning and can provide personalized services according to user preferences;with the help of Service Robot Laboratory in Shandong University,the feasibility of the method proposed in this paper is verified.