Recognition Of Social Anxiety Through Analysis Of Galvanic Skin Response Signal

Social anxiety, a highly prevalent and complicated emotional state, is defined as the negative emotional state induced by the excessive concern on the expected or real negative judgment from others when a person is in a real or even an imaginary social situation. In general, social anxiety includes the psychological(such as tension, nervousness or apprehension) and physical symptoms(such as chest congestion, sweating, muscle tension, etc.). Analysis of the anxiety-specific physiological patterns can help recognize the state of social anxiety, and the results can be applied in the field of healthcare, lie detection, telemedicine, etc. In this thesis, social anxiety of 22 subjects are induced through a verified Trier Social Stress Protocol, and the galvanic skin response during social anxiety is analyzed to find out anxiety-specific skin conductance patterns.First, the thesis gave a detailed introduction on the experimental protocol and characters of social anxiety state. Then, according to the experimental protocol the study devised an experiment to elicit public speaking fear, one of the most representative subtypes of social anxiety, under controlled laboratory conditions. Moreover, the paper verified the successful arousal of social anxiety state in the experiment by combining the self-reported anxiety of the subjects and the anxiety score given by the 21 audiences. Two channels of GSR signal, the GSR at the shoulder zone and the GSR at the fingers, were acquired during the experiment. Furthermore, the paper applied the feature optimization algorithm of ReliefF, the classification algorithm of k-Nearest Neighbor(kNN) and Error Back-Propagation(BP) to recognize the social anxiety state. The results illustrate that recognition of social anxiety through GSR at the shoulder zone is better than that at the fingers.In a word, the current work selected anxiety-specific skin conductance features and classified GSR data of three different levels of anxiety, namely non-anxiety, low anxiety and high anxiety. The current work is beneficial for the quantitative modeling of anxietyrelated skin conductance response.