Automatic detection of nausea using bio-signals during immersion in a virtual reality environment

VR (Virtual Reality) systems have been widely used for various purposes. However, during people's immersion in a virtual environment it is commonly reported that simulation sickness can occur, and it prevents us from utilizing a VR environment for wider purposes. We constructed a controlled VR environment for analyzing the change of bio-signals during VR immersion, where subjects were requested to find trash cans in the virtual environment within five minutes. Each subject's various bio-signals, which were EEGs from 5 different locations, vertical EOG, lead I ECG, fingertip skin temperature, photoplethysmogram, and skin conductance level, were measured during experiments. We analyzed and compared the signals, and we found out that the characteristics of 28 signals during nausea were statistically different from when the subjects were at rest, or during the first 30 seconds after the immersion was started. We parameterized these characteristics and established 12 principal components using principal component analysis in order to reduce the redundancy in those parameters, and constructed an artificial neural network with those principal components. Using the network we constructed, we could partially detect nausea in real time.