An algorithm for detecting startle state based on physiological signals

To construct safe workspaces or daily life environments, it is important to develop a technology for automatically recording dangerous environmental situations. Based on increased finger skin conductance and decreased instantaneous heart (pulse) rate, we developed an algorithm for detecting the startle state of humans when encountering dangerous situations. The parameter values in the algorithm were optimized by tuning them with experimental results obtained in a virtual reality system, in which dangerous scenarios were presented to 21 subjects. The detection sensitivity of the optimized algorithm was 100% during a static standing condition and also while cycling. The detection specificity attained 90% and 87% for the static and exercising conditions, respectively. The optimized algorithm was applied to a real world situation such as car driving, and relatively high sensitivity of 87% and specificity of 80% were obtained.