Correlation Analysis Between Environmental Sound and Human Emotion

In daily life, as people are most exposed to the surrounding environment, accordingly humans are greatly affected as their emotion depending on the surrounding visual and spatial information. In this paper, we propose an analysis method of how the sound information such as amplitude and frequency in the surrounding environment can affect to human emotions by adopting sound features among the visual and spatial information. For the experiments, a total of 1,500 video clips of surrounding environment are acquired by 15 subjects using the camera built in smartphone. Also, the subjective evaluation were performed after taking the video. Two features such as amplitude and frequency of sound data were extracted. Then, we designed a fully connected SVR inference networks in which the data were divided into two sets such as training and test. The extracted two dimensional features were SVR trained by corresponding with the subjective evaluation scores for pleasant and arousal levels. As a result, we confirmed that the estimated two-dimensional emotions were similar with the subjective evaluated ones in which the errors were about the pleasant level (−1 ~ +1) of 0.27 and the arousal level (−1 ~ +1) of 0.32, respectively.