Clothing classification with smart phones

Abstract

Human thermal comfort is significantly dependent on thermal insulation of clothing [3]. Therefore, classifying types of clothing a user is wearing plays an important role in enhancing human thermal comfort. In our work, we investigated different combinations of feature extraction methods and machine learning algorithms for clothing classification. We conducted our study using temperature and humidity data collected from smartphones in various contexts (inside and outside a pocket) and with different clothing types. We found that using six largest coefficients returned from Discrete Wavelet Transform with Support Vector Machines learning algorithm, we can achieve an accuracy of up to 0:71.