In-clothing Climate Sensing to Predict Comfort in Each Wearing Situation

Abstract

In order to improve the quality of life, people wear comfortable clothing according to the time and occasion. However, some TPOs may sacrifice some comfort. In this study, we propose a method to estimate comfort for each TPO in order to realize clothing recommendation that recommends enough or acceptable comfortable clothing for the most important TPO of the day.

In this study, we hypothesize the following from the relationship between the outside temperature, the climate inside the clothes, and comfort. We hypothesized that the degree of comfort in a certain TPO is estimated from the following variables: temperature and humidity inside clothing and outside-air, their first-order and second-order differentiation, difference between inside-clothing and outside-air temperatures and humidities, and activity level in the TPO.

The proposed method collects temperature and humidity inside and outside clothing, and heart rate and activity levels through wearable sensors as well as comfort in each TPO through questionnaire for building prediction models. As a result of the experiment, the accuracy rate was better for XGBoost in 4 out of 5 subjects than RandomForest model. It is possible to use it for clothing recommendation considering the comfort of each TPO based on temperature and humidity in clothing, outside climate, the sensor data, and activity level in TPO.