A Study On Time Series Analysis Of Environmental Data For Predicting Emotional Conditions

Emotion estimation technology has been attracting attention from the viewpoints of work style reform, class support, and car driving support. Conventional emotion estimation techniques were based on facial expressions, biometric data and language. Besides, image- and voice-based emotion estimation methods have privacy issues. On the other hand, our previous study had shown the effectiveness of emotion prediction from environmental data. However, time-series prediction for changes in emotion has not been developed yet. In this study, we aim to build a system to predict human emotional conditions in time series using environmental data. This study showed that a deep-learning approach was effective in predicting the time series of emotional data from environmental data. We also found that the accuracy of time series emotion prediction depends on the specific time period in a day.