“I KNOW HOW YOU FEEL” – PREDICTING EMOTIONS FROM SENSORS FOR ASSISTED PEDELEC EXPERIENCES IN SMART CITIES

Abstract

Emotions are one of the manner humans use to indicate how they feel about a particular event, place or things. To date there is no consensus about the correlation of measured data to an unambiguously defined emotional state. The selection of parameters, their weight and range, which derive at an emotion, are not clearly defined. Especially, if measurements took place outdoors and during a physical activity. This work is based on previous work and focuses on the parameters and methods to classify measured data to an emotional state. We took a closer look to the values, defined ranges for parameters and performed further pre-processing steps. Furthermore, we revised the assignment of an emotion, analyzed the parameter weights and their correlation. Moreover, we compared our previous approach with further Machine Learning (ML) methods. The results are in line with previous work, however, indicate the need for more and heterogeneous data to endorse the outcome. Further results from the parameter analysis suggest an importance of the skin conductance level (SCL) depending on the method used.